# Stage 1 - Reading the spec (27 Opus agents)

## spec-map (synthesis)

1. Listed /tmp/claude-1000/-home-cfcosta-Notes/f3cf6639-a28c-403a-997d-fc0414dcd4e4/scratchpad/digest/ and confirmed 8 r-files and 18 d-files. 2. Read all eight rules-side digests in full: r1-front-intro.md (Highlights, ToC, complete subject-to-rule index, General introduction incl. history table, 13-entry bibliography, "9 sections / 53 rules" statement), r2-sec1-words.md (rules 1.1-1.14, 22 technical-noun categories, 4 technical-verb categories), r3-sec2-3.md (rules 2.1-2.2, 3.1-3.7), r4-sec4-sentences.md (rules 4.1-4.5), r5-sec5-6.md (rules 5.1-5.5, 6.1-6.6), r6-sec7-8.md (rules 7.1-7.3, 8.1-8.7), r7-sec9-practices.md (rules 9.1-9.4 plus GR-1..GR-8 with GR-7 skipped), r8-dict-intro.md (dictionary structure, 875/1274 claim, 39 recurring errors, 208 approved verbs, missing flowchart). 3. Did not read the d*-dict files, per instructions; used the 18 reader summaries supplied in the task for coverage, counts, and anomalies. 4. Verified the rule inventory: 14+2+7+5+5+6+3+7+4 = 53 numbered rules across 9 sections, matching the General introduction's own claim in r1; GR-1..GR-8 tracked separately as non-rules. 5. Summed dictionary counts with a script: approved 879, not-approved 1319, total 2198 vs spec 2149; deltas +0.46%/+3.53%/+2.28%. 6. Checked all 17 alphabetical junctions between d-batches for continuity (appear->APPLICABLE, brilliance->bring, ..., till->TILT, utilize->valid): contiguous, one junction (impact at d08/d09) flagged for verification. 7. Wrote map.md with the six requested sections: standard overview; complete Part 1 rule index per section; Part 2 structure, per-file count table, and spec-delta analysis; cross-cutting concept location table; the three recorded locations of the skipped inclusive-language material (GR-7 at page 1-9-12, its Highlights line, its SRI row) plus the two retained external citations; and a per-digest anomaly rollup ordered by fact-check priority.

gaps: Dictionary counts: approved 879 vs spec 875 (+4, +0.46%, within tolerance); not-approved 1319 vs spec 1274 (+45, +3.53%, exceeds ~2%). NO letter range looks short - coverage is contiguous A ("A (art)") through Z ("zero") with clean junctions at all 17 batch boundaries, and the deviation is an OVERAGE, not a shortfall. Most plausible cause: readers counted one entry per headword+part-of-speech row (substitute adj/n/v=3, detail n/v, whichever adj/pron, yet conj/adv, MIDDLE adj/n, GROUND n/v, etc.) while the official 875/1274 likely counts distinct words; fact-check should normalize multi-PoS rows to confirm. One boundary to verify: d08/d09 at "impact" (d08 ends "up to impact" at page 2-1-I2; d09 starts at "impact (v)" on 2-1-I3) - confirm no double-count or dropped row. Genuinely missing content (graphics absent from pdftotext, must be read from the PDF images): the "How to select words correctly" flowchart on dictionary page 2-0-16 (r8) and the Rule 3.6 passive-to-active diagram (r3). Reconstructed-symbol spots to confirm against the PDF: degree/ohm/arc-second symbols in Rule 1.5 category 9 (r2) and the ohm symbol in Rule 8.6 example (r6). Lightly garbled source prose: d12 "preferable (adj)" STE/Non-STE example (alternative RECOMMEND unaffected); d13 STE-example-column typos ("FO THE COMPONENT", dropped WITH/WITH AN); d07 "file (v)" unclosed parenthesis reconstructed. Source-internal wording variances captured, not errors by readers: Rule 3.6 if/when, Rule 4.3 text/texts, Rule 8.6 element-6 wording and Structural/Structure Repair Manual, SRI footer date 2024-04-30 vs 2025-01-15. Highlights contain no dictionary word-by-word change log (only per-rule Part 1 changes) - absent from source batch, not an extraction gap.

## r1-front-intro.md - Full batch covered (source lines 1-827; pdf pages HI-1..HI-6, TOC-1, SRI-1..SRI-4, general introduction i-viii). Content

1. Read the entire source file (827 lines) top to bottom in two passes (lines 1-420 and 420-827), covering Highlights, TOC, Subject-to-rule index, and the general introduction. 2. Ran a targeted grep for inclusive-language material: found 4 hits — 2 STE topic pointers (Highlights GR-7 change line 255; Subject-to-rule index row line 385) skipped per the user constraint and recorded as a single skip note; 2 external bibliographic citations (lines 805, 821) kept inside the required Reference documents list. 3. Extracted Issue 9 Highlights GENERAL/structural changes (source lines 6-257): general preliminary-page changes (page status/date, terminology alignment with ISO, Help added, title/subtitle/copyright/TOC/SRI/general-introduction revisions) and the per-rule change summary for Part 1 Sections 1-9 plus GR-1 through GR-8; flagged the substantive structural items (new Categories 1.5-21/1.5-22, restructured 1.12 Category 3 + new Category 4, Rule 2.3 removed and moved to become Rule 4.5, tables removed/revised in Section 3, new GR-7 and GR-8, 8.6 sub-item changes incl. revised word count in 8.6.2, "technical noun" replacing "technical name"); omitted the boilerplate and the absent dictionary change log. 4. Extracted the Table of contents (lines 267-306): front-matter items with page prefixes, Part 1 Sections 1-9 (1-1-1..1-9-1), Part 2 Dictionary (Introduction 2-0-3, Word list 2-1-A1). 5. Reproduced the COMPLETE Subject-to-rule index (lines 329-482, pages SRI-1..SRI-4) as a two-column table with the `#`-convention note; the only omission is the inclusive-language row (skip note). 6. Extracted the FULL General introduction (lines 493-826, pages i-viii): every Q&A subsection (what STE is; when/why developed; purpose; which words are available incl. the "technical nouns"/"technical verbs" definition, one-meaning/one-part-of-speech principle, synonym-selection, American English/Merriam-Webster basis; abbreviations; text formatting; units of measurement; use alone; oral communication; teaching English; technical translation with NMT/LLM note; training & authoring tools with the accreditation exception), the History release table (1986 first release through Issue 9), Maintenance (STEMG/STEST), Distribution (free of charge; unauthorized distribution prohibited), Change forms + the change-form template field list, Acknowledgments/dedication (four named chairpersons with dates), Guide to the writing rules (9 sections / 53 rules; American English spelling/punctuation; the Non-STE vs STE example convention with the valve example), Help, and the complete 13-entry Reference documents bibliography. 7. Determined the batch has no rule bodies, so set items=0 and ruleIndex=[]; wrote the digest to the out path and recorded all OCR/format anomalies and the deliberate inclusive-language omission.

**Anomalies:** No numbered rule bodies exist in this batch, so 0 rules were extracted and ruleIndex is empty (this is expected — the batch is front matter + general introduction; rule statements/titles are not given, only per-rule change descriptions). OCR/typo artifacts, reproduced corrected with the source form noted here: (a) source line 337 shows "1,5" (comma) in the Abbreviations index row — clearly rule 1.5, reproduced as "1.5"; (b) source line 382 shows "Hyphens ... 2.2, 8.2. 8.7" (stray period) — reproduced as "2.2, 8.2, 8.7". Footer date inconsistency: Subject-to-rule index pages SRI-1..SRI-4 carry footer date "2024-04-30" while every other page carries "2025-01-15" (likely a leftover; the document's official Issue 9 date is 2025-01-15). Running headers/footers and one "Blank Page" marker (back of TOC-1) were ignored as instructed. One deliberate omission (not an artifact): the "Inclusive language → 9 – GR-7" subject-index row and the GR-7 Highlights line were skipped per the user constraint and folded into a single skip note. Two reference-documents citations touching gender-sensitive / disability-inclusive communication (source lines 805, 821) were retained as part of the required complete bibliography, since they are external-document citations rather than STE's own inclusive-language guidance.

## r2-sec1-words.md - Part 1, Section 1 (Words), complete — rules 1.1 through 1.14, pdf pages 45-62 (internal page numbers 1-1-1 through 1-1-1

1. Read the entire source file (829 lines) top to bottom, confirming it spans internal pages 1-1-1 through 1-1-18 (Section 1, Words). 2. Extracted the opening "Summary of the rules" (page 1-1-1) into the section introduction, capturing the seven topic groupings (Which words can you use?; Part of speech; Approved meaning; Forms of verbs and adjectives; Technical nouns; Technical verbs; Spelling) and the ISO 1087:2019 definition of "subject fields". 3. Extracted Rule 1.1 with its technical-noun/technical-verb definitions and the use/engine/ream examples. 4. Extracted Rule 1.2 (part of speech) with the Test, Dim, clean, acceptable, and operable examples plus the not-in-dictionary lookup Help. 5. Extracted Rule 1.3 (approved meanings) with the follow/obey example. 6. Extracted Rule 1.4 (approved forms) including the REMOVE verb-form table and the SLOW adjective-form table and the more/most note. 7. Extracted Rule 1.5 in full, reproducing all 22 technical-noun categories with their complete example word lists (flagging the category 9 symbol loss and the category 7 stray period). 8. Extracted Rules 1.6 through 1.11 (technical-noun usage: unapproved-as-noun, noun-not-as-verb, company-approved, short/<=three-words, no regional/slang/jargon, consistent naming) each with statement, detail, and example pairs — preserving the "not more than three words" limit verbatim in 1.9. 9. Extracted Rule 1.12 in full, reproducing all 4 technical-verb categories with sub-groups a-f and complete example verb lists, plus the "prefer approved dictionary verbs" guidance and the detect/ream/clamp/grease/wire/enter/respond examples. 10. Extracted Rules 1.13 (verb-not-as-noun, past-participle-as-adjective, plate example) and 1.14 (American English spelling, fibre/colour examples, Rule 8.6 cross-reference). 11. Recorded cross-references (Section 3 from 1.12; Rule 8.6 from 1.14) and verified no inclusive-language content exists in this batch, so the USER CONSTRAINT skip-line was not needed. 12. Compiled the ruleIndex of all 14 rule numbers and their exact statement-titles.

**Anomalies:** Two source artifacts worth a fact-checker's attention, both from pdftotext, neither affecting rule text: (1) In Rule 1.5, category 9 (Numbers/units/time), the special symbols for "degree" (°), "ohm" (Ω), and the arc-"second" (") were dropped and rendered as empty parentheses in the source; I reconstructed them from standard usage and flagged this inline in the digest. (2) In Rule 1.5, category 7, the source printed "geology. geophysics" with a stray period where a comma belongs; I normalized it to a comma. No entries are truncated at the batch boundaries (the batch is a self-contained section, 1-1-1 to 1-1-18, ending on a Blank Page). Page headers/footers and the "Blank Page" marker were ignored as instructed.

## r3-sec2-3.md - Part 1 Section 2 (Multi-word nouns), rules 2.1-2.2 (PDF pages 1-2-1 to 1-2-4); and Part 1 Section 3 (Verbs), rules 3.1-3

1. Read the entire source file (593 lines) top to bottom in one pass. 2. Scanned the whole batch for any inclusive-language material per the user constraint; found none, so nothing was skipped on that basis. 3. Extracted the Section 2 (Multi-word nouns) heading, its "Summary of the rules" block, and the section introduction (definition of multi-word nouns, head-noun-is-last-word point, ambiguity rationale). 4. Extracted Rule 2.1 (three-word maximum) with its detail, preposition guidance, and all four Non-STE/STE example pairs including the parenthetical word counts. 5. Extracted Rule 2.2 (write in full, then Method 1 shorter form / Method 2 hyphens) with the full two-method detail, the "hyphenated words count as one word" and "no more than three words with hyphens" limits, the approved-abbreviation and inward-outward-valve conditions, and representative examples. 6. Extracted the Section 3 (Verbs) heading and its three-part "Summary of the rules" (Verb forms/tenses; Active voice; How to describe an action). 7. Extracted Rule 3.1 (verb forms from the dictionary) with REMOVE/GIVE examples and the Part 2 dictionary-intro Help note. 8. Extracted Rule 3.2 (six approved verb forms/tenses), reconstructing the interleaved Adjust/Give table and copying the four not-approved constructions (present perfect, past perfect, present/past progressive, other complex constructions). 9. Extracted Rule 3.3 (past participle as adjective — before a noun; after to be/become/stay; not-passive note; (adj) entries like permitted/damaged) with examples. 10. Extracted Rule 3.4 (no auxiliary verbs for complex constructions) with the "have" prohibition and all five Non-STE/STE pairs. 11. Extracted Rule 3.5 ("-ing" only as technical noun or modifier) with the four grammatical functions, permitted heading/modifier uses, and the complete approved "-ing" word list (nouns lighting/opening/routing/servicing; adjectives mating/missing/remaining; pronoun something; preposition during). 12. Extracted Rule 3.6 (active voice; passive only if agent unknown) including active/passive definitions, the "by whom or by what?" test, the transmission/data-corruption example, and all four passive-to-active methods with their example pairs plus the procedural-vs-descriptive examples; noted the summary-vs-heading wording variance. 13. Extracted Rule 3.7 (use an approved verb, not a noun/other part of speech) with the ohmmeter, removal, and check-the-battery examples. 14. Built the ruleIndex covering rules 2.1, 2.2, 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7 and wrote the digest to the target path.

**Anomalies:** Rule 3.6 is worded slightly differently in the section summary ("...passive voice only when the agent is unknown") versus the rule heading body ("...only if the agent is unknown"); both variants recorded. The passive-to-active "Method 1" references "the diagram that follows," but that diagram is an image absent from the pdftotext output (only its caption/example survives). The Rule 3.2 six-form verb table was column-interleaved in the source; it was reconstructed into a markdown table. Expected repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Part 1 - Writing rules", "Page 1-x-y") and one "Blank Page" marker (line 586) were present and ignored. No inclusive-language material found, so nothing was skipped on that basis.

## r4-sec4-sentences.md - Part 1, Section 4 (Sentences), pdf pages 1-4-1 through 1-4-10. Full section: the "Summary of the rules" front block, the

1. Read the entire source file (415 lines) top to bottom in one pass; confirmed it contains only Section 4 (Sentences), pdf 1-4-1 to 1-4-10. 2. Extracted the section-head "Summary of the rules" block, preserving its four sub-headings (Short sentences and clear sentence structures; Vertical lists; Connecting words and connecting phrases; Articles and demonstrative adjectives) and the rules listed under each. 3. Summarized the Section 4 introduction (basic principle of STE = short sentences/clear structures; cross-refs to procedures=section 5, descriptive text=section 6; Section 4 = general rules for both). 4. Extracted Rule 4.1 (Write short and clear sentences.) with its imperative-for-procedures / one-topic-per-descriptive-sentence / be-accurate details and the cover-assembly, side-stay, and no-leaks/cure-time example pairs. 5. Extracted Rule 4.2 (Do not omit words or use contractions...) verbatim statement plus all five omission categories (noun, verb, subject, article, contraction) each with its STE/Non-STE example including the potable-water WARNING. 6. Extracted Rule 4.3 (Use a vertical list for complex text.) with the full 8-point construction checklist, the 4-way item-marker sub-list, the Help box, and the five post-example conditions (no mixing writing types, DO NOT per safety item, connect each item to the lead-in, keep one level/flatten nesting, verb-but-not-full-sentence = no period); reproduced the wheel-assembly, RAT-procedure, and CAUTION safety examples and noted the remaining examples. 7. Extracted Rule 4.4 (connecting words/phrases) with approved connectors ("and," "but," "then," "thus," plus phrases "as a result"/"at the same time"), demonstrative-adjective connectors, and all STE examples. 8. Extracted Rule 4.5 (articles/demonstrative adjectives) with the no-article-for-general-concepts detail, the article-in-a-series ambiguity rule (the two "new O-rings" contrasting sentences), the alphanumeric-identifier/proper-noun rule (circuit breaker 36L7), and the "Practical English Usage" Help box. 9. Built ruleIndex for 4.1–4.5, checked for the USER CONSTRAINT topic (no inclusive-language content found), and recorded the 4.3 title plural/singular discrepancy in anomalies.

**Anomalies:** Wording mismatch: the section-head "Summary of the rules" prints Rule 4.3 as "Use a vertical list for complex texts." (plural), while the rule box on pdf 1-4-4 prints "Use a vertical list for complex text." (singular). I used the rule-box (singular) form as the canonical title and flagged the summary variant in the digest. Standard page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Part 1 - Writing rules", "Page 1-4-x") and one "Blank Page" marker (pdf 1-4-10) were present and ignored. The 4-item "how to identify list items" bullet block (dash/bullet/letter/number) and several multi-line WARNING/CAUTION examples wrapped across lines but reassembled cleanly with no garbling or truncation at the batch boundaries. No inclusive-language material present, so no skip line was needed.

## r5-sec5-6.md - Part 1 Section 5 (Procedural writing) rules 5.1-5.5, and Part 1 Section 6 (Descriptive writing) rules 6.1-6.6. Includes 

1. Read the entire source file (659 lines) top to bottom in one pass; confirmed it spans Section 5 and Section 6 only. 2. Scanned the whole batch for any "inclusive language" material per the user constraint — none present, so no skip line was recorded. 3. Extracted Section 5's "Summary of the rules" header (grouped: Sentences 5.1-5.2, Verbs in procedures 5.3, Descriptive statements 5.4, Notes 5.5). 4. Extracted Rule 5.1 (20-word procedural max) with its 4 example pairs and both Help boxes (Section 8 pointer; 25-word note-sentence limit). 5. Extracted Rule 5.2 (one instruction per sentence; same-time exception; multi-sentence work-step cases) with all example sets. 6. Extracted Rule 5.3 (imperative form; the three ambiguities; the "must" safety exception) with examples. 7. Extracted Rule 5.4 (condition-first + comma; comma-placement meaning shift) with the CSD examples. 8. Extracted Rule 5.5 (notes = information only; 25-word/sentence; no imperative; safety-instruction and limit exceptions) plus the "How to use notes correctly" test and its light-bulb Help. 9. Extracted Section 6's "Summary of the rules" header and the "Content structure" introduction (imperative not permitted; three kinds of descriptive writing). 10. Extracted Rule 6.1 (give information gradually, one subject per sentence) with the full Instrument Landing System Non-STE/STE pair. 11. Extracted Rule 6.2 (key words/phrases, connecting words "and/but/then/thus", connecting phrases "as a result/at the same time") including the sentence-by-sentence linking walkthrough. 12. Extracted Rule 6.3 (25-word descriptive max vs 20-word procedural) with the smartphone example pair. 13. Extracted Rule 6.4 (paragraphs + topic sentence) with the Lightning Strike example. 14. Extracted Rule 6.5 (one topic per paragraph; topic-sentence-as-outline) with the two-topic ILS breakdown. 15. Extracted Rule 6.6 (max six sentences per paragraph) with the full fuel-manifold example and its per-paragraph sentence counts. 16. Added a cross-batch numeric-limits summary (20/25/25/6 words-sentences) and wrote the digest to the target path. 17. Built ruleIndex covering all 11 rules (5.1-5.5, 6.1-6.6).

**Anomalies:** No garbled or truncated text at batch boundaries: the file opens cleanly on the Section 5 header and closes cleanly on the 1-6-8 "Blank Page". Minor source inconsistencies preserved as-is: (1) in Rule 6.2's comparison tables, sentence 1 is quoted two ways — "The Instrument Landing System (the system) on the aircraft shows data..." and later "The Instrument Landing System in the aircraft shows data..."; (2) the Rule 5.1 measurement appears as "(0.24 inches)" in the Non-STE line but "(0.24 in)" in the STE line (intentional, demonstrating STE abbreviation). Four-column sentence-linking tables were interleaved by pdftotext but were fully reconstructable. "Blank Page" markers at pages 1-5-8 and 1-6-8 and repeated page headers/footers ignored. No inclusive-language topic found, so the skip line was not needed.

## r6-sec7-8.md - Part 1 Section 7 (Safety instructions) Rules 7.1-7.3 and Section 8 (Punctuation and word count) Rules 8.1-8.7, PDF pages

1. Read the entire source file r6-sections7-8-safety-punctuation.txt (lines 1-541) top to bottom. 2. Scanned the whole batch for inclusive-language material per the user constraint — none present, so no skip line added. 3. Extracted Section 7 header, Summary of the rules, and the Definitions introduction (warning vs caution definitions + the ISO 45001 / ANSI Z535 / ISO 3864 references and light-bulb help), lines 5-33. 4. Extracted Rule 7.1 (statement, uppercase-formatting note, risk-analysis decision list, warning/caution examples, oxygen-tubes Non-STE vs STE pair), lines 46-89. 5. Extracted Rule 7.2 (command-first vs condition-first, four safety examples), lines 98-119. 6. Extracted Rule 7.3 (explain the risk/result, reusing the rule 7.2 examples), lines 130-149. 7. Extracted Section 8 header, Summary of the rules (Rules 8.1-8.7 with the parentheses and word-count sublists), and the Punctuation introduction with its five reference books, lines 160-199. 8. Extracted Rule 8.1 (no semicolon; two Non-STE/STE pairs), lines 208-221. 9. Extracted Rule 8.2 (hyphens; all five categories with their full example word lists; dash-vs-hyphen light-bulb), lines 225-251. 10. Extracted Rule 8.3 (seven permitted parentheses uses, each with STE examples), lines 260-316. 11. Extracted Rule 8.4 (colon as period in vertical lists; copied the 20-word procedural / 25-word descriptive limits exactly; the fire-extinguisher list example with per-line word counts), lines 321-339. 12. Extracted Rule 8.5 (parenthetical text = one word; EMER example and three identifier examples with counts), lines 350-369. 13. Extracted Rule 8.6 (all seven element types — numbers, numbers+units, abbreviations, alphanumeric identifiers, quoted text, titles/headings/placards/labels, proper nouns — with every example and its word count, plus the paragraph-number light-bulb), lines 373-517. 14. Extracted Rule 8.7 (hyphenated words = one word; four examples with counts), lines 521-534. 15. Recorded the three anomalies and populated ruleIndex with all 10 rule numbers and exact titles.

**Anomalies:** Three source/pdftotext artifacts noted in the digest: (1) Rule 8.6 §2 example "The resistance must be 10" (marked 5 words) lost its ohm symbol (Ω) in pdftotext extraction; reconstructed from the paired "...10 ohms." example. (2) Rule 8.6 §6 example text says "Structural Repair Manual" while its parenthetical explanation says "Structure Repair Manual" — inconsistency in the original. (3) Rule 8.6 sixth element is worded "Titles, headings, placards, and labels" in the Summary-of-rules but "Titles, headings, and text on placards and labels" in the rule body — both captured. Standard page headers/footers were ignored. Otherwise clean; no truncation at batch boundaries.

## r7-sec9-practices.md - Part 1, Section 9 (Writing practices), PDF pages 1-9-1 through 1-9-14. Full section: the "Summary of the rules" box, 4 n

1. Read the entire source file r7-section9-practices.txt (616 lines) top to bottom in one pass. 2. Parsed the "Summary of the rules" box (lines 7-24) and reproduced its grouped listing of Rules 9.1-9.4 and GR-1 through GR-8. 3. Extracted Rule 9.1 (lines 27-199): statement, the word-for-word-replacement concept, the four reasons a different construction is needed, the "How to use a different sentence construction" guidance, the worked crack-length table example (24-word/23-word counts and passive-voice notes copied exactly), and the STE primary-objective statement; captured 4 example pairs. 4. Extracted Rule 9.2 (lines 202-294): restricted-meaning principle, approved-part-of-speech principle, dual-function "flush" note; captured all 7 restricted-meaning examples, 3 wrong-part-of-speech examples, and 2 "flush" examples. 5. Extracted Rule 9.3 (lines 298-336): phrasal-verb definition, "put out" concrete/abstract illustration, the not-permitted normative statement, and the note that only a few phrasal verbs (e.g. "put on", "come on") are approved; captured 2 example pairs. 6. Extracted Rule 9.4 (lines 339-392): consistent-style statement, the two-valid-sentences illustration, the 8-step procedure example with its three inconsistency categories, and the descriptive-writing note. 7. Extracted the General recommendations intro (lines 395-397) noting GRs are not STE rules. 8. Extracted GR-1 (that), GR-2 (with), GR-3 (pronouns), GR-4 (this), GR-5 (false friends), GR-6 (Latin abbreviations) with all their WRITE/Do-not-write example pairs (lines 400-556). 9. Located GR-7 Inclusive language (lines 567-582, PDF 1-9-12) and, per the user constraint, recorded only a one-line skip note without extracting content. 10. Extracted GR-8 Possessive form (lines 585-597). 11. Wrote the digest to the target path and compiled ruleIndex covering all 4 rules and all 8 GRs.

**Anomalies:** Clean overall. The 4-column example procedure under Rule 9.4 (steps 1-8, non-STE vs STE) and the two-column crack-length tables under Rule 9.1 were column-interleaved in the pdftotext output but reconstructed correctly. The degree symbol rendered fine ("500 °C"). Repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Part 1 - Writing Rules/rules", "Page 1-9-x") and one "Blank Page" marker (page 1-9-14) were ignored as artifacts. No truncation at batch boundaries — the section starts and ends cleanly. One deliberate omission: GR-7 (Inclusive language) content skipped per user instruction (only its existence/location recorded).

## r8-dict-intro.md - Part 2 Dictionary Introduction, pdf pages 131-148 = dictionary pages 2-0-3 through 2-0-19. Extracted: General intro (875

1) Read the entire source file (718 lines, pages 2-0-3 to 2-0-19) top to bottom. 2) Ran a grep for inclusive-language terms; only match was 'inclusion' (dictionary inclusion, page 2-0-14) — confirmed no inclusive-language topic, so no skip line required. 3) Extracted the General introduction (page 2-0-3): approved-word count 875, non-approved 1274, TN/TV markers, no antonyms, American English / Merriam-Webster spelling. 4) Extracted the 4-column Guide (page 2-0-3). 5) Extracted Column 1 (pages 2-0-4 to 2-0-8): bold/UPPERCASE-vs-lowercase convention, the eight parts of speech (n, v, adj, adv, pron, art, prep, conj) with full definitions, and Forms of approved words for nouns (singular; plural of countables permitted; ref Rule 1.5 technical nouns), verbs (permitted forms only; transitive/intransitive; ref Part 1 Sec 3 and Rule 1.12 technical verbs), adjectives, and adverbs. 6) Extracted the four verb types (pages 2-0-5 to 2-0-7): Regular (ADAPT), Irregular (GIVE), Irregular auxiliary (BE, no other forms), Defective modal (CAN/COULD, WILL — no other forms; help 'do not use COULD for possibility'). 7) Extracted Column 2 (pages 2-0-9 to 2-0-10): approved meaning vs. ALTERNATIVES, part-of-speech alignment, TN/TV alternatives (uncovered→COVER (TN)), phrase alternatives (simultaneously→AT THE SAME TIME). 8) Extracted the four Help categories (pages 2-0-11 to 2-0-13): recommendations (PUSH), restricted meaning (ABOUT), single-context (SWALLOW safety only), important info (BE, already). 9) Extracted Column 3 (page 2-0-14) and Column 4 (page 2-0-15) descriptions with examples (ABOVE→MORE THAN). 10) Extracted the "How to select words correctly" intro (page 2-0-16) and flagged the missing flowchart graphic. 11) Reproduced the COMPLETE recurring-errors table verbatim as a two-column map (39 entries across pages 2-0-17 and 2-0-18 'continued'). 12) De-interleaved and reproduced the COMPLETE approved-verbs list (page 2-0-19), counting 208 verbs by letter group and verifying the per-letter subtotals sum to 208. 13) Wrote everything to the digest file with anomaly notes.

**Anomalies:** 1) The "How to select words correctly" FLOWCHART (page 2-0-16) is a graphic and is NOT in the pdftotext output — only its two-sentence intro is present; the decision-flow steps could not be extracted because they are absent from the source, not omitted by me. 2) The approved-verbs list was printed as a 5-column interleaved table; I de-interleaved and re-sorted it into a single alphabetical list (208 verbs; letters J, N, Q, V, X, Y, Z have no verbs). 3) No numbered rules in this batch, so ruleIndex is empty; the 'items' count = 39 recurring-error mappings + 208 approved verbs. 4) No inclusive-language content present (the only 'inclusion' hit on page 2-0-14 is about adding examples to the dictionary), so no skip line was needed. 5) Page headers/footers and one 'Blank Page' marker (page 2-0-20) were ignored as instructed. Text is otherwise clean with no garbling or truncation at batch boundaries.

## d01-dict.md - Letter A, headwords "A" through "appear"; dictionary pages 2-1-A1 through 2-1-A16 (pdf 149-164).

1. Read the entire source file (811 lines) top to bottom and confirmed via wc -l that no content was cut off. 2. Confirmed the batch is dictionary letter A, pages 2-1-A1 to 2-1-A16, by tracking the page footers (A1 line 46 through A16 line 810). 3. Verified the opening entry "A (art)" (line 5) starts a fresh headword, so nothing was skipped as a prior-batch continuation. 4. Walked every leftmost-column headword row and classified by case: UPPERCASE = APPROVED, lowercase = NOT APPROVED; captured (TN)/(TV) markers. 5. Extracted APPROVED entries A1-A16: A, ABOUT, ABOVE, ABRASIVE, ABSORB (2 meanings), ACCEPT, ACCESS, ACCIDENT, ACCIDENTAL, ACCIDENTALLY, ACCURATE, ACCURATELY, ACROSS, ACTIVATE, ACTIVE, ADAPT, ADD, ADJACENT, ADJACENT TO, ADJUST, ADJUSTABLE, ADJUSTMENT, AFT (adj), AFT (adv), AFTER (conj), AFTER (prep), AFT OF, AGAIN, AGAINST, AGENT, AGREE, AID, AIRBORNE, AIRFLOW, ALIGN, ALL, ALMOST, ALONG, ALSO, ALTERNATIVE (adj), ALTERNATIVE (n), ALTHOUGH, ALWAYS, AN, ANALOG, AND, ANGULAR, APART (48 entries), preserving "for other meanings, use" alternatives for ABOUT/ABOVE/ACTIVATE in brackets. 6. Extracted NOT-APPROVED entries in order from abaft through appear (83 entries: abaft, abandon, abate, ability, able, abnormal, abnormality, abrupt, absence, absent, absolutely, absorption, abundant, abut, accelerate, acceptable, acceptance, accessible, accommodate, accomplish, according to, account for, accumulate, accumulation, accuracy, achieve, acquire, acrid, act, action, activity, actuate, actuation, addition, additional, adequate, adhere, adhesion, adjoining, admit, adopt, advance (n), advance (v), adverse, advisable, advise, affect, aggravate, agitate, agitation, ahead, aid, aim, air-dry, alert, alignment, allocate, allow, allowable, already, alter, alteration, alternate (adj), alternate (v), alternately, alternatively, amend, amendment, among, amount, analyze, anchor, ancillary, annotation, annunciate, another, anticlockwise, any (adj), any (pron), anytime, aperture, apparent, appear), keeping approved-alternative UPPERCASE words with their (pos)/(TN) tags and compressing editorial-note-only guidance to square-bracket hints. 7. Cross-checked the count by re-scanning all headword lines: 48 approved + 83 not approved = 131 items, within the expected 100-140 range. 8. Wrote the two-section digest to the output path.

**Anomalies:** No garbled/unreconstructible text. The 4-column layout interleaves (STE example and Non-STE example wrap across lines), but every entry was cleanly reassembled. No mid-entry truncation at either batch boundary: the batch opens with a full "A (art)" entry (no orphaned continuation from a prior batch) and closes with a complete "appear (v)" entry ending at POSSIBLE (adj) on page A16. No inclusive-language material appears in this dictionary batch (that topic lives in Part 1 writing rules, not the Part 2 word list), so the user's skip instruction did not apply here. Note: "ABSORB (v)" is one headword with two numbered approved meanings (rendered as 2 lines but counted as 1 approved entry). A few NOT-APPROVED entries give phrase/guidance alternatives rather than single words: "act (v)" and "any (adj)" give only an editorial note (compressed to square-bracket hints); "already (adv)", "adverse (adj)", "affect (v)", "accomplish (v)" mix word alternatives with a note.

## d02-dict.md - Part 2 Dictionary word list, pdf pages 165-180 = dictionary pages 2-1-A17 through 2-1-B8. Headwords from APPLICABLE thro

1. Read the entire source file (788 lines) top to bottom, ignoring repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Part 2 - Dictionary", "Page 2-1-Axx/Bx"). 2. Verified the batch opens on a fresh headword APPLICABLE (adj) (line 5) with no orphan continuation text, confirming nothing rolled over from the prior batch. 3. Transcribed the A-tail approved (UPPERCASE) headwords in order: APPLICABLE, APPLY (2 meanings), APPROVAL, APPROVED, APPROXIMATE, APPROXIMATELY, AREA, ARM (2 meanings), AROUND, ARROW, AS (prep), AS...AS, ASSEMBLE, ASSEMBLY, AT, ATTACH, AUTHORITY, AUTOMATIC, AUTOMATICALLY, AUXILIARY (2 meanings), AVAILABLE, AWAY FROM, AXIAL, AXIALLY. 4. Transcribed the A-tail not-approved (lowercase) headwords with their approved alternatives: application, appoint, approach, appropriate, approve, arise, arrange, arrangement, arrest, arrive, article, as (conj), ascertain, ask, assess, assessment, assign, assist, assistance, associated, assume, assure, as to, asymmetric, attachment, attain, attempt (n), attempt (v), attention, audible, augment, aurally, authentic, authorized, avert, avoid, aware. 5. Transcribed the B-section approved headwords: BACK (adv), BAD, BALANCE, BARE, BASIC, BE (2 meanings), BECAUSE, BECAUSE OF, BECOME, BEFORE (conj), BEFORE (prep), BEHIND, BELOW, BEND (n), BEND (v), BETWEEN, BLEED, BLOCKAGE, BLOCKED, BLOW, BLUNT, BOND (n, 2 meanings), BOND (v, 2 meanings), BOTTOM (adj), BOTTOM (n), BREAK, BREATHE, BRIGHT, BRIGHTLY. 6. Transcribed the B-section not-approved headwords: back (adj), back (n), back and forth, back off, backup, backward, badly, bank, bared, base (n), base (v), begin, beginning, below (adv), beneath, beside, beware, beyond, big, bind, blank, blend, block, blot, blunted, bolt, both, bottom (v), bounds, brace, branch, breadth, break (n), breakdown, brilliance. 7. Preserved multi-part-of-speech and multi-meaning splits on separate lines (attempt n/v; base n/v; back adj/adv/n; BEFORE conj/prep; BEND n/v; BOND n/v; BOTTOM adj/n) and kept TN/TV markers (CONFIGURATION (TN), CAP (TN), COVER (TN), BANK (TN), THE TWO (TN), BOLT/BOLTS (TN)). 8. Checked the whole batch for inclusive-language content per the user constraint (none present). 9. Tallied 53 approved and 72 not-approved headword entries = 125 items, and wrote the two-section digest to the out path.

**Anomalies:** No garbled or unreconstructable column text; the 4-column layout reassembled cleanly throughout. No inclusive-language material appears in this batch (that topic is not present here), so nothing was skipped for the user constraint. The file starts on a clean headword (APPLICABLE (adj), line 5) with no continuation text, so no mid-entry fragment carried over from the previous batch was skipped. One normalization: the bolt (v) entry printed its technical-name markers as "[TN]" in square brackets (WITH A BOLT [TN] OR BOLTS [TN]); I rendered them as (TN) to match the dictionary's usual convention. Bracketed notes I added to a few approved entries (AROUND, AS...AS, BAD, BE, BECOME, BELOW) capture the source's cross-reference/caution sub-notes; the "big" alternative includes a source sub-instruction rendered as [give an accurate value if possible].

## d03-dict.md - Part 2 Dictionary word list, alphabetical run from "bring" through "comply" (letters B tail and C), pdf pages 181-196 = 

1. Read the entire source file /tmp/.../spec/d03-dict-p181-196.txt (776 lines) top to bottom and confirmed line count with wc -l. 2. Verified the batch opens at a real headword "bring (v)" (line 5) rather than a headless continuation, so nothing from the prior batch needed skipping. 3. Walked the B-tail entries on page 2-1-B9/B10: bring, briskly, broadcast, brush, brush on, BUBBLE, build(n), build(v), bump, BURN, burst, BUT, butt, BY, by means of, BYPASS. 4. Walked C entries on pages 2-1-C1 through C4: CALCULATE, calculation, CALIBRATE, CALIBRATION, call, CAN, CANCEL, CANNOT, cap, capability, capable, CAPACITY, care, CAREFUL, CAREFULLY, carry, carry out, case, case(in case of), CATCH, categorize, CAUSE(n), CAUSE(v), caution, cautiously, cavity, cease. 5. Walked pages 2-1-C5 through C8: center, centralize, certain, certify, chafe, chafed, chance, CHANGE(n), CHANGE(v), channel, characteristic, CHARGE, CHECK(n), check(v), CHEMICAL, CHEMICALLY, chip(n), chip(v), chock, choice, choose, circa, circle, CIRCULAR, circulate, clamp, classification, CLEAN(adj), CLEAN(v), cleanliness. 6. Walked pages 2-1-C9 through C11: CLEAR, clear(v), CLEARANCE, CLEARLY, CLICK, click(v), clip, CLOCKWISE, CLOGGED, close(adj), CLOSE(v), closely, close to, coat(n), coat(v), coating, cock, CODE, code(v), COIL, coil(v), coincide, coincident, COLD, collapse, COLLECT, COLOR, color code, colored. 7. Walked pages 2-1-C12 through C14 to the batch end: combine, combustible, COME, COME ON, commence, commencement, common, communicate, COMMUNICATION, COMPARE, comparison, COMPATIBLE, compile, complete(adj), COMPLETE(v), COMPLETED(adj), completely, completion, complicated, comply. 8. Classified each headword by case: UPPERCASE = approved (column-2 meaning transcribed terse), lowercase = not approved (column-2 ALTERNATIVES listed with pos and TN/TV markers preserved). 9. For approved verbs, retained the conjugation forms from the headword cell; for words with two numbered senses (BURN, CANCEL, CLEAR, CLOSE) repeated the headword on a second line per the format. 10. Compressed guidance/construction notes into short square-bracket hints and cross-references into bracketed "for other meanings use ..." hints. 11. Counted 45 approved + 77 not-approved = 122, and wrote the two-section digest.

**Anomalies:** No garbled or unreconstructable text; the 4-column layout reassembled cleanly for every entry. File starts cleanly at a headword ("bring (v)") on page 2-1-B9, so no mid-entry continuation from the previous batch was skipped. No inclusive-language material appears anywhere in this batch (pure alphabetical word list), so no skip line was added. Transcription notes kept faithful to source: (1) "common (adj)" gives its approved alternative as "SAME (adv)" in the source — an unusual pos tag for "same" but transcribed verbatim; (2) three not-approved entries have multi-word uppercase phrase alternatives rather than single headwords — "compile (v)" -> MAKE A LIST, "combine (v)" -> PUT TOGETHER, "complicated (adj)" -> NOT EASY — kept verbatim; (3) "comply (v)" carries a special note that it is allowed in law and regulations (refer to rule 1.12), captured as a bracketed hint alongside its alternative OBEY (v). Approved-entry "for other meanings, use:" cross-references (CATCH->COLLECT, CHANGE(v)->REPLACE, CHARGE->FILL/PRESSURIZE, CLEAR(adj)->TRANSPARENT) and "no other verb forms" notes (CAN, COME, COME ON) were folded into the approved meaning lines as bracketed hints.

## d04-dict.md - Part 2 Dictionary word list, pdf pages 197-212 (internal page labels 2-1-C15 through 2-1-D6). Headwords COMPONENT throug

1. Read the entire source file d04-dict-p197-212.txt (789 lines) top to bottom and confirmed length with wc -l; stripped repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Page 2-1-Cxx/Dxx", "Part 2 - Dictionary") and column-header rows. 2. Verified the batch opens with a complete headword (COMPONENT (n), page 2-1-C15) — no mid-entry continuation to skip. 3. Worked page 2-1-C15 to C18 (COMPONENT..CONTACT): classified UPPERCASE headwords as APPROVED (COMPONENT, COMPRESS, CONCENTRATION, CONCENTRIC, CONDENSATION, CONDITION [2 senses], CONNECT, CONNECTION, CONSTANT, CONSTANTLY, CONTACT) and lowercase as NOT APPROVED with their column-2 alternatives (comprehensive, comprise, compulsory, compute, concentrate, concern, conclusion, concurrently, conduct, confine to, confirm, conflict, conform, conformance, conformity, consecutive, consecutively, consequence, consider, considerable, consist of, consistent, construct, consult). 4. Worked pages 2-1-C19 to C24 (CONTAIN..CYCLE): APPROVED (CONTAIN, CONTAINER, CONTAMINATION, CONTENTS, CONTINUE, CONTINUOUS, CONTINUOUSLY, CONTOUR, CONTROL(n), CONTROL(v), COOL(adj), COPY(n), CORNER, CORRECT(adj), CORRECT(v), CORRECTION, CORRECTLY, COUNT, COUNTERCLOCKWISE, CURVE(n), CUT [2 senses], CYCLE); NOT APPROVED (contaminant, contaminated, contiguous, conventional, converse, convert, convey, cool(v), coordinate, copy(v), correctness, correspond, corresponding, corrode, corrosive, couple, cover, crack, create, critical, cross, curvature, curve(v), cycle(v)). 5. Worked pages 2-1-D1 to D6 (DAMAGE..denote): APPROVED (DAMAGE, DAMAGED, DANGEROUS, DATA [2 senses], DATE, DEACTIVATE, DECISION, DECREASE, DEDICATED, DE-ENERGIZE, DEEP, DEFECTIVE, DEFLATE, DEFUEL); NOT APPROVED (damage(v), damp, dampen, danger, dangerously, deadly, deaerate, dearm, decay, decelerate, decide, decontamination, decrease(n), decrement, define, deflect, deflection, deformed, defueling, degrease, de-ice, delay(n), delay(v), delete, delicate, deliver, delivery, demand, denote). 6. Preserved (TN)/(TV) and (pos) markers exactly as printed in alternatives (e.g. ONE (TN) AFTER THE OTHER, CORROSION (TN), CRACK (TN), COVER (TN), DEFORMATION (TN), GREASE (TN), ICE (TN), AIR (TN), RADIUS (TN)); kept multi-word approved-phrase alternatives verbatim in uppercase (AT THE SAME TIME, VERY IMPORTANT, MAKE MOIST, EASILY DAMAGED, MAKE A CURVE, REMOVE CONTAMINATION). 7. Captured verb inflection lists on approved verbs (e.g. COMPRESS/COMPRESSES/COMPRESSED, CUT/CUTS/CUT/CUT) and the DEEP comparative/superlative form. 8. Scanned the whole batch for any inclusive-language topic — none present. 9. Counted 47 approved + 77 not-approved = 124 entries and wrote the two-section digest in encounter order.

**Anomalies:** No continuation/orphan entry at the start: the batch opens cleanly with the full headword COMPONENT (n) after the column-header row, so nothing was skipped as belonging to a previous batch. No inclusive-language material is present in this batch (nothing skipped per the user constraint). No garbled or unreconstructable column text encountered; the 4-column layout reassembled cleanly throughout. Three approved headwords carry two numbered senses each and are recorded as two meaning-lines but count as one entry each: CONDITION (n), CUT (v), DATA (n). Embedded usage notes preserved in [square brackets] on their lines: CONDITION (use IF for "on the condition that"), DATA (singular allowed), CONTACT (other meanings use TOUCH), DEACTIVATE (other meanings use STOP), danger (technical-noun-for-safety-instruction, refer to section 7). CONTROL and CORRECT each appear as two separate headword entries (noun + verb, and adj + verb respectively), counted separately.

## d05-dict.md - Part 2 Dictionary word list, pdf pages 213-228 (footer pages 2-1-D7 through 2-1-D19 and 2-1-E1 through 2-1-E2). Headword

1. Read the entire source file (755 lines) top to bottom, identifying the 4-column table layout (Word+POS | Approved meaning/ALTERNATIVES | STE EXAMPLE | Non-STE example) and stripping page headers/footers and the "Blank Page" marker at 2-1-D20. 2. Worked the D7-D9 span (lines 5-141): transcribed dent through detrimental — approved DEPLOY, DEPTH, DETERIORATION; the rest lowercase/not-approved with their alternatives (e.g. deposit -> PARTICLE/CONTAMINATION; detach -> DISCONNECT/DISENGAGE/REMOVE; determine -> FIND/GIVE/SELECT/CALCULATE). 3. Worked D9-D11 (lines 152-243): develop through dilute — approved DEVICE, DIAGONALLY, DIFFERENCE, DIFFERENT, DIFFERENTLY, DIGITAL, DIGITALLY. 4. Worked D12 (lines 254-292): dim through DIRTY — approved DIM, DIMENSION, DIMLY, DIRECTION, DIRECTLY, DIRTY. 5. Worked D13 (lines 303-344): disappear through discrepancy — approved DISARM (2 meanings), DISASSEMBLE, DISCARD, DISCONNECT. 6. Worked D13-D15 (lines 355-443): DISENGAGE through disturb — approved DISENGAGE, DISPLAY (n), DISTANCE; captured disposition -> LOCATION/POSITION and distribute -> APPLY/SUPPLY. 7. Worked D15-D17 (lines 454-533): DIVIDE through draw — approved DIVIDE (2 meanings), DO (2 meanings incl. auxiliary a/b), DOWN (adj/adv/prep), DOWNSTREAM (adj/adv), DOWNSTREAM OF (prep), DRAIN. 8. Worked D17-D19 (lines 544-636): dress through dust — approved DRINK, DROP, DRY (adj/v), DURING; captured drift's DRIFT (TN) qualifier and the DRINK SWALLOW-note. 9. Crossed the D20 blank page into E1-E2 (lines 662-748): EACH through efficient — approved EACH (adj/pron), EACH OTHER, EASILY, EASY, EAT, EDGE, EFFECT; not-approved early, earth (n/v), ease (n/v), effect, effective, efficacious, efficient. 10. Split entries into APPROVED (uppercase headwords, 45) and NOT APPROVED (lowercase headwords, 74), wrote the two-section digest in encounter order, and cross-checked the start (fresh at "dent") and end (complete at "efficient") for boundary truncation; scanned the whole batch for inclusive-language material (none found).

**Anomalies:** No garbled or unreconstructable text — all 4-column rows reassembled cleanly. No inclusive-language material present in this batch (nothing skipped for that reason). No mid-entry continuation at the start: the file opens at the complete headword "dent (v)", so no leading entry was carried over from the previous batch. The file ends with "efficient (adj)" fully transcribed (its Non-STE example ends before the page 2-1-E2 footer), so no trailing entry was truncated — nothing skipped. Notes on counting: DISARM, DIVIDE, and DO each have two numbered meanings (rendered as two lines each in the digest) but are counted as one approved entry each. DOWN (adj/adv/prep), DOWNSTREAM (adj/adv) plus DOWNSTREAM OF (prep), EACH (adj/pron), and DRY (adj/v) are counted as separate approved entries per distinct part-of-speech row; likewise detail (n/v), earth (n/v), and ease (n/v) on the not-approved side. Two lowercase entries had reworded (non-lexical) alternatives which I hinted in [brackets]: disappear -> VIEW (n) [go out of view], droop -> MOVE (v) [move down]. ease (v) alternatives are approved-word phrases kept uppercase (CAREFULLY REMOVE; CAREFULLY MOVE). drift's second alternative carries a qualifier: REMOVE (v) (WITH A DRIFT (TN)). Approved entries DRINK and EAT carry a "no other verb forms / for safety instructions use SWALLOW (v)" note preserved parenthetically.

## d06-dict.md - Part 2 Dictionary word list, "E" section from "efficiently" through "exude" (pdf pages 229-244; dictionary pages 2-1-E3 

1. Read the entire source file /tmp/.../spec/d06-dict-p229-244.txt (785 lines) top to bottom, identifying it as the "E" run of the Part 2 Dictionary (pages 2-1-E3 to 2-1-E18). 2. Confirmed the batch opens with a complete headword ("efficiently (adv)", line 5) rather than continuation text, so no leading entry belongs to the previous batch; confirmed it closes with a complete entry ("exude (v)", lines 776-778). 3. Walked each 4-column row, using headword case to classify: UPPERCASE = STE-approved (column-2 meaning), lowercase = not approved (column-2 alternatives). Ignored repeated page headers/footers ("ASD-STE100...", "Issue 9", "2025-01-15", "Page 2-1-Exx", "Part 2 - Dictionary") and the per-page column-title rows. 4. Extracted 36 APPROVED headwords in order: EJECT, ELECTRIC, ELECTRICAL, ELECTRICALLY, ELECTROMAGNETIC, ELECTROMAGNETICALLY, ELECTRONIC, ELECTRONICALLY, EMERGENCY, EMPTY, END, ENERGIZE, ENGAGE, ENTRANCE, ENTRY, EQUAL, EQUALLY, EQUIPMENT, EQUIVALENT(adj), EQUIVALENT(n), ERASE, ERROR, ESTIMATE, EXAMINE, EXAMPLE, EXHAUST, EXIT, EXPAND, EXPIRED, EXPLOSION, EXPLOSIVE, EXTEND, EXTENSION (2 senses), EXTERNAL, EXTERNALLY, EXTINGUISH — each with its terse column-2 definition. 5. Extracted 79 NOT-APPROVED headwords in order (efficiently, efflux, effort, either[adj/conj/pron], ejection, elapse, eliminate, emit, employ, empty, enable, encircle, enclosed, encounter, end, enforce, engagement, enlarge, enough, ensue, ensure, enter, entire, entirely, equal, equip, equipped, erect, erratic, escape, establish, estimate, evaluate, evaluation, evaporate, even, evenly, event, eventually, every, evidence[n/v], evident, exact, exactly, examination, exceed, except, exception, exceptional, exceptionally, excess[adj/n], excessive, excessively, exchange, exclude, excluding, execute, exempt, exercise, exert, exhale, exhaust, exhaustive, exist, expect, expel, explain, expose, extent, exterior, extra, extract, extreme, extremely, exude), each mapped to its column-2 approved alternatives in listed order, preserving (pos)/(TN) markers and compressing the two rewording-only entries (except/exception) to bracketed hints. 6. Wrote the two-section digest (## APPROVED then ## NOT APPROVED), one line per meaning/alternative set, repeating headwords with multiple pos or senses. 7. Re-counted via awk: 37 approved lines (= 36 distinct entries, EXTENSION doubled) and 79 not-approved lines; reported approved=36, unapproved=79, items=115.

**Anomalies:** No garbled or unreconstructable column text; the 4-column layout reassembled cleanly throughout. Notes: (1) No mid-entry continuation at the top — the batch starts cleanly with the complete entry "efficiently (adv)"; nothing was skipped at either boundary, and the file ends cleanly with "exude (v)". (2) Two headwords were hyphen-split across lines in the source and reassembled: "ELECTROMAG-NETIC" -> ELECTROMAGNETIC (adj), and "ELECTROMAG-NETICALLY" -> ELECTROMAGNETICALLY (adv); one STE example also carried the split "ELECTROMAG-NETICALLY". (3) EXTENSION (n) carries two numbered senses (1. the action when something extends; 2. the result when something extends) — rendered as 2 lines in the digest but counted as 1 approved entry (hence 37 approved lines vs 36 distinct approved entries). (4) "except (prep)" and "exception (n)" give no word substitute but the instruction "To replace this word, use a different construction. Be accurate." — compressed to [use a different construction; be accurate]. (5) The alternative FORCE for "effort (n)" is marked (TN) technical name; ANALYSIS for evaluate/evaluation is (TN); "ONE OF THE TWO" and "ONE OR THE OTHER" for "either" use (TN) markers on ONE/TWO. (6) EXTINGUISH (v) is approved (To stop burning) but carries a usage note that for lights or lamps GO OFF / BE OFF must be used instead — captured inline in the meaning. No inclusive-language material appears in this batch.

## d07-dict.md - Part 2 Dictionary word list, PDF pages 245-260 (dictionary pages 2-1-F1 through 2-1-G2). Headwords from "fabricate" thro

1. Read the entire source file (806 lines) top to bottom, spanning dictionary pages 2-1-F1 to 2-1-G2. 2. Confirmed the first line of content is a fresh headword (fabricate, F) and not a continuation, so no leading entry was dropped; confirmed the final entry (GENERAL, G) is complete. 3. Ran a grep for inclusive-language terms across the batch — no matches — so the user constraint did not trigger. 4. Classified each headword by case: UPPERCASE = APPROVED (column-2 meaning captured), lowercase = NOT APPROVED (column-2 uppercase alternatives captured). 5. Extracted APPROVED entries in source order across letters F–G: FACE, FACT, FALL, FAR, FAST, FEATHER, FEEL, FILL, FILTERED, FIND, FINISH, FIRE(n), FIRE(v), FIRST, FIT, FLAME, FLAMMABLE, FLANGE, FLASH, FLAT, FLEXIBLE, FLIGHT, FLOW(n), FLOW(v), FLUID, FLUSH(adj), FLUSH(v), FOLD, FOLLOW, FOR, FOR EXAMPLE, FORWARD(adj), FORWARD(adv), FORWARD OF, FREE, FREELY, FREEZE, FREQUENT, FREQUENTLY, FROM, FRONT(adj), FRONT(n), FULL, FULLY, FUNCTION, GAS, GENERAL = 47 approved. 6. Extracted NOT-APPROVED entries in source order F–G (fabricate … gauge), reassembling multi-alternative sets across wrapped column text (e.g., faulty->4 alts, fix->4 alts, foul->4 alts) = 75 unapproved. 7. Preserved verb inflections and adjective comparatives in headword brackets, and TN/TV markers exactly as given. 8. Wrote the two-section digest (## APPROVED / ## NOT APPROVED) and tallied 47 + 75 = 122 items.

**Anomalies:** No garbled/unreconstructable text — the 4-column layout was clean and every entry reassembled confidently. No entry was skipped: the file starts with the complete headword "fabricate (v)" (line 5, not a mid-entry continuation from the previous batch), and ends with the complete entry "GENERAL (adj)" (last full entry before the page 2-1-G2 footer). No inclusive-language material is present in this batch (grep for inclusive/gender/he-she/his-her returned no matches), so nothing was skipped on that basis. Minor notes captured as bracketed hints rather than dropped: FACE's "technical noun for body part" note; FOLLOW's "for other meanings use OBEY" note; FREE's compound-adjective prohibition; and the technical-noun condition on lowercase "failure (n)" (approved as TN only when it means a performance error / loss of serviceability). "file (v)" alternative had an unclosed paren in the source ("REMOVE (v) (WITH A FILE (TN)") reconstructed as "REMOVE (v) (WITH A FILE (TN))".

## d08-dict.md - Part 2 Dictionary word list, letters G (mid, from "generally" on pages 2-1-G3 onward) through H (complete) into I (from 

1. Read the entire source file /tmp/.../spec/d08-dict-p261-276.txt top to bottom (765 lines) and confirmed it ends at the Page 2-1-I2 footer with no further content. 2. Confirmed the first row (line 5, "generally (adv)") is a complete headword entry, so no mid-entry continuation from the prior batch needed skipping. 3. Scanned the whole batch for any "inclusive language" topic material; found none (it is a plain A-Z word-list, letters G/H/I only), so no content was withheld under the user constraint. 4. Extracted letter-G entries from "generally" through "gush" (pages 2-1-G3 to 2-1-G8), classifying UPPERCASE headwords (GET, GIVE, GLOSSY, GO, GO OFF, GOOD, GRADUALLY, GROOVE, GROUND n, GROUND v, GROUP) as APPROVED and lowercase headwords as NOT APPROVED with their column-2 alternatives. 5. Extracted all letter-H entries "halt" through "hydraulically" (pages 2-1-H1 to 2-1-H7), including approved HANG, HARD, HAVE, HEAD, HEAR, HEAT, HEAVY, HEIGHT, HELP, HERE, HIGH, HIT, HOLD (2 senses), HOLE, HORIZONTAL, HORIZONTALLY, HOT, HOW, HYDRAULIC, HYDRAULICALLY; noted the 2-1-H8 Blank Page. 6. Extracted letter-I entries "identical" through "impact" (pages 2-1-I1 to 2-1-I2): approved IDENTIFICATION, IDENTIFY (2 senses + FIND cross-ref), IF, IGNORE, ILLUMINATE (with COME ON/BE ON restriction), IMMEDIATELY; unapproved identical, idle, ignite, immediate, immerse, immobilize, impact. 7. Reassembled wrapped/interleaved 4-column rows for multi-alternative entries (glitch, hinder, handle, hesitation, guide, grab, immerse, immobilize, great, heat, glue, harden, harm, harmful, harmonize, haul) and preserved TN/TV markers and rule references. 8. Wrote the two-section digest in encounter order and tallied 37 approved + 79 not-approved = 116 entries.

**Anomalies:** No garbled/unreconstructable text; the 4-column interleaving (esp. immobilize, immerse, hazard/hazardous, ILLUMINATE, IDENTIFY notes) was reassembled confidently. The batch begins partway through letter G, but starts cleanly at a full headword "generally (adv)" (line 5) - no headwordless continuation from the prior batch to skip. G1-G2 entries (letters before "generally") were in a previous batch. One "Blank Page" marker at page 2-1-H8 (line 653). No inclusive-language material appears anywhere in this batch (straight A-Z word list only), so nothing was skipped on that account. Repeated page headers/footers and column-header rows ignored as instructed. Phrasal/reworded alternatives kept as given: "GO AWAY", "GO DOWN", "GO INTO", "MOVE OFF", "MORE THAN", "VERY MUCH", "TIGHTEN ... WITH YOUR HAND", "PUT FULLY INTO", "PREVENT MOVEMENT"; "have to" -> [use a verb in the imperative form] compressed as a bracketed hint. Approved entries GET/ILLUMINATE/IDENTIFY carry usage-restriction notes preserved in brackets; HOLD and IDENTIFY each have two approved meanings (two lines each). GROUND appears as two separate approved entries (n and v). Source notes (rule refs, safety-sign/safety-regulation flags, "if possible be accurate") preserved as brief bracketed annotations.

## d09-dict.md - Part 2 Dictionary word list, entries "impact" through "just" — I-section from impact onward (pages 2-1-I3 to 2-1-I16) pl

1. Read the entire source file (lines 1-717) top to bottom, covering dictionary pages 2-1-I3 through 2-1-J2. 2. Stripped repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Part 2 - Dictionary", "Page 2-1-Ix/Jx") and the "Blank Page" marker on page 2-1-J2. 3. Confirmed the batch opens at a complete headword (impact (v), line 5) — no continuation entry to skip. 4. Reassembled each 4-column row (Word/pos | Approved meaning or ALTERNATIVES | STE EXAMPLE | Non-STE example) across wrapped lines, classifying UPPERCASE headwords as APPROVED and lowercase as NOT APPROVED. 5. Extracted APPROVED entries I-section: IMPORTANT, IN(adv), IN(prep), IN FRONT OF, IN PROGRESS, INBOARD(adj), INBOARD(adv), INBOARD OF, INCIDENT, INCLUDE, INCORRECT, INCORRECTLY, INCREASE, INCREMENT, INDEPENDENTLY, INDICATION, INFLATE, INFORMATION, INITIAL, INITIALLY, INJURY, INNER, INPUT, INSPECTION, INSTALL, INSTALLATION, INSTRUCTION, INSTRUMENT, INSULATION, INTENSITY, INTERCHANGE, INTERCHANGEABLE, INTERFACE, INTERMITTENT, INTERMITTENTLY, INTERNAL, INTERNALLY, INTERVAL, INTO, INVESTIGATION, IRREGULAR, IRREGULARLY, ISOLATE, IT, ITEM, ITS; and J-section: JOINT (47 total), recording verb inflection lists and the two numbered sub-senses of INTO and IT. 6. Extracted NOT APPROVED entries with their approved alternatives (65 total): impact, impair, implement, implementation, impossible, impregnate, impression, improper, improperly, improve, impurity, inactive, inadvertent, inadvertently, incline, including, incompatible, incomplete, incorporate, incorporation, increase, incur, independent, indicate, induce, ineffective, inflammable, inform, ingest, ingress, inhale, inhibit, initiate, inject, injection, injure, inoperative, insert, inside(adj), inside(n), inside(prep), inspect, instead, instead of, insufficient, insure, intact, integral, interconnect, interfere, interference, interior, intermediate, interrupt, intersect, investigate, inward(adj), inward(adv), irregularity, isolation, jack, jam, job, join, just. 7. Preserved alternative markers exactly (TN technical name, e.g. JACK/INTERFERENCE/INTERSECTION/DEFECT; pos tags), compressed non-word usage notes to bracketed hints. 8. Counted 47 approved + 65 not approved = 112 items and wrote the two-section digest to the out path.

**Anomalies:** No garbled or unreconstructable text — column interleaving (e.g. impurity line 76, independent line 245, intermediate lines 554-558) was cleanly reassembled. File begins at a full headword "impact (v)" with no mid-entry continuation from a previous batch, so nothing was skipped at the boundary. No inclusive-language topic present, so the skip rule did not apply. Two entries carry usage notes rather than word alternatives, compressed to bracketed hints: interference (n) [can be a technical noun or part of a technical noun] and intermediate (adj) BETWEEN (prep) [give the limits]. impurity's first alternative "UNWANTED MATERIAL" has no (pos) marker in the source and is transcribed as given. Both interfere (v) and interference (n) are distinct lowercase entries with near-identical alternative sets; both retained. jam (v) -> FREELY (adv) is unusual but is exactly as printed.

## d10-dict.md - Part 2 Dictionary word list, letters K-L-M. Headwords KEEP through MAXIMUM (n), spanning pages 2-1-K1 to 2-1-M4 (pdf 293

1. Read the source file in full (736 lines, top to bottom) covering dictionary pages 2-1-K1 through 2-1-M4; confirmed EOF at line 736 with wc -l. 2. Stripped repeated headers/footers (ASD-STE100 title, Issue 9, 2025-01-15, Page 2-1-x markers) and two "Blank Page" markers. 3. K section (lines 5-32): extracted 7 entries - approved KEEP, KILL, KNOW; not-approved key, kind, kink, knock. Verified KEEP starts as a full headword (no continuation from prior batch). 4. L section (lines 58-531): extracted 82 entries. Approved (44): LAMINATED, LAMINATION, LARGE, LAST(adj), LAST(adv), LATCH, LATERAL, LATERALLY, LAYER, LEAK(n), LEAKAGE, LEFT, LENGTH (two numbered senses), LESS, LET, LEVEL(adj), LEVEL(n), LIFE, LIFT, LIGHT, LIGHTING, LIGHTLY, LIMIT, LINEAR, LINEARLY, LIQUID(adj), LIQUID(n), LISTEN, LIVE, LOCAL, LOCALLY, LOCATION, LOCK, LONG, LONGITUDINAL, LONGITUDINALLY, LOOK, LOOP, LOOSE, LOOSELY, LOOSEN, LOW, LOWER, LUBRICATE. Not-approved (38): label, lack, land, last(v), later(adj), later(adv), lay, lead, leak(v), least(adv), least(at least), leave, left-hand, legible, lengthen, lethal, level(v), liberal, liberally, lie, light(v), limitation, limited, link, list, little(adj), little(a little)(adj), little(a little)(adv), load, locate, log, long(as long as), long(no longer), loop(v), lose, lost, lowermost, lukewarm. 5. M section (lines 542-732): extracted 30 entries. Approved (18): MAGNETIC, MAGNETICALLY, MAINTENANCE, MAKE, MAKE SURE, MALFUNCTION, MANDATORY, MANUAL(adj), MANUAL(n), MANUALLY, MANY, MARK(n), MASS, MATERIAL, MATING, MATT, MAXIMUM(adj), MAXIMUM(n). Not-approved (12): main, maintain, major, make certain, maladjusted, manipulate, manufacture, mark(v), mask, match, mate, matter. 6. Reassembled column-wrapped meanings and multi-alternative redirect lists per entry; separated adj/adv/n homographs (LAST, LEVEL, LIQUID, MANUAL, MAXIMUM) into distinct lines. 7. Tallied 65 approved + 54 not-approved = 119 items; cross-checked the sum against per-letter counts (K=7, L=82, M=30).

**Anomalies:** No garbled/unreconstructable text - all 4-column entries reassembled cleanly. No mid-entry continuation to skip: the file opens with the complete headword KEEP (v), not a fragment from the previous batch. No inclusive-language topic appears in this K/L/M batch. Two "Blank Page" markers and repeated page headers/footers ignored as instructed. Note: LIGHT (adj), LIVE (adj), and LOOSE (adj) are hybrid entries - approved for a specific meaning but redirect to alternatives (THIN/SMALL, ENERGIZE, TENSION) for other senses; the redirect is captured as a bracketed hint on the approved line. Alternatives like NOT SUFFICIENT, SMALL QUANTITY, INCORRECTLY ADJUSTED, COME ON, LET … STAY, NOT FIND, NOT … AT THIS TIME, UNTIL … NOT are standard STE multi-word constructions given in uppercase in column 2 and kept verbatim.

## d11-dict.md - Part 2 Dictionary word list, pages 2-1-M5 through 2-1-O6. Letters M through O (partial O): headwords from "may" through 

1. Read the full source file (791 lines) top to bottom, spanning dictionary pages 2-1-M5 to 2-1-O6. 2. Ran a grep for inclusive-language keywords (inclusive/gender/he-she/his-her) across the file - zero matches, so the user constraint did not apply to this batch; all content extracted. 3. Stripped repeating page artifacts (header "ASD-STE100 Simplified Technical English", column headers, "Issue 9 / 2025-01-15 / Part 2 - Dictionary / Page 2-1-x" footers) and reassembled each wrapped 4-column row into a single entry. 4. Letter M (pages M5-M10): extracted may through MUST - approved MEASURE, MECHANICAL, MECHANICALLY, MECHANISM, MEDICAL, MELT, METHOD, MIDDLE(adj/n), MINIMUM(adj/n), MINUS, MISSING, MIX, MIXTURE, MOBILE, MODE, MODERATE, MODERATELY, MOIST, MOMENTARILY, MONITOR, MOOR, MORE(adj/adv), MOST(adj/adv), MOVABLE, MOVE, MOVEMENT, MUCH(adj/adv), MULTIPLY, MUST; not-approved may, medium, meet, mention, mesh, meteorological, mid, mild, minimize, minor, miscellaneous, misrigged, mix(n), modify, moisten, momentary, motion, motor, mount. 5. Letter N (pages N1-N4): approved NAME, NEAR(adj/prep), NECESSARY, NEUTRAL, NEW, NEXT, NO(adj/adv), NOISE, NONE, NOSE, NOT, NOTCH, NUMBER; not-approved nearly, need, neither(adj/pron), neutralize, never, next(adv), next to, noisy, nor, normal, normally, note, notify, now. 6. Letter O (pages O1-O6): approved OBEY, OBJECT, OCCUR, OF, OFF(adj/adv/prep), ON(adj/adv/prep), ONE, ONLY(adj/adv), ONTO, OPEN(adj/v), OPENING, OPERATE, OPERATION, OPPOSITE(adj/prep), OPTIONAL, OR, OTHER(adj/pron), OUT(adj/adv), OUT OF; not-approved observe, obstruct, obstructed, obstruction, obtain, often, oil, oil-free, old, omit, once(adv/conj), opaque, operable, operational, opportunity, option, order(n/v), orient, origin, original, originate, otherwise. 7. Wrote APPROVED section (77 headword+POS entries, one meaning per line with multi-meaning rows split) and NOT APPROVED section (58 entries with their uppercase approved alternatives / TN/TV markers) to the digest, entries in source order. 8. Counted approved=77 and unapproved=58 separately, items=135, within the expected 100-140 range.

**Anomalies:** No garbled text; all 4-column rows reassembled confidently. First entry "may (v)" begins with its own headword (not a mid-entry continuation), so no boundary entry was skipped. Three entries needed judgment calls, all preserved faithfully: (1) MULTIPLY (v) is APPROVED for "to use multiplication to get a result" but carries a "For other meanings, use: INCREASE (v)" note - captured in a [bracket] on the approved line. (2) ONE (pron) and operational (adj) carry technical-noun notes - captured in [brackets]. (3) opportunity (n) gives guidance ("If it is possible, be accurate.") rather than a word substitution - compressed to [rewrite to be specific/accurate, e.g. AT THE NEXT CHECK]. Short approved-word phrases used verbatim as alternatives (DO NOT, AT THIS TIME, NOT TRANSPARENT, INCORRECTLY ADJUSTED, TWO (TN) … NOT, ONE (TN) TIME, IF …NOT, DO NOT DO) were kept uppercase as given, not bracketed. Counting note: entries counted by distinct headword+part-of-speech row (e.g. MIDDLE adj and MIDDLE n = 2; neither adj and neither pron = 2); multi-meaning rows like MIX/NEUTRAL/NUMBER/OFF/ON/OPEN/OPPOSITE/OUT OF appear as two lines each but count as one entry.

## d12-dict.md - Part 2 Dictionary word list, pdf pages 325-340 (dictionary pages 2-1-O7 through 2-1-P12). Entries alphabetically from OU

1. Read the entire source file d12-dict-p325-340.txt (779 lines) top to bottom; confirmed length via wc -l and that content ends at PROBLEM (n) on page 2-1-P12. 2. Verified the leftmost/headword column and part-of-speech to classify APPROVED (UPPERCASE) vs NOT APPROVED (lowercase); identified TN/TV and (TN) markers. 3. Worked the "O" tail (2-1-O7 to O9): OUTBOARD/OUTBOARD OF/OUTDOORS/OUTER/OUTPUT/OVERBOARD/OVERLAP/OVERRIDE approved; outline, outside(adj/adv/n/prep), outwards, over, overfill, overfull, overhaul, overheat, overlap(v), overtighten not approved; noted 2-1-O10 Blank Page. 4. Worked the "P" range (2-1-P1 to P12): reassembled wrapped column-2 text for each entry (e.g., PART's two senses, POINT (n) and POINT (v) two senses each, POSITION, PIECE, PERFORMANCE, PERSON/PERSONNEL, PRESSURIZATION). 5. Captured every not-approved entry's full alternative set in order (e.g., over -> ABOVE/ALONG/ON/MORE THAN + technical-noun note; pass -> GO/THROUGH/GIVE; passage -> FLOW/HOLE/OPENING; poor -> UNSATISFACTORY/DEFECTIVE/UNSERVICEABLE; priority -> IMPORTANT/BEFORE/FIRST; probable -> VERY POSSIBLE/RISK), preserving TN markers (overhaul->OVERHAUL (TN), preload->LOAD (TN), preserve->PRESERVATION (TN), placard->PUT (v) (A PLACARD [TN]), plot->MAKE (v) (A PLOT [TN]), probe->FIND (v) (WITH A PROBE [TN])). 6. Compressed embedded guidance notes to bracketed hints ([use a more accurate verb] for perform; [give the value] for periodically; LAST-adj note on PREVIOUS). 7. Counted entries by distinct headword+part-of-speech (multi-sense words counted once): 47 approved, 77 not approved, total 124. 8. Scanned for any inclusive-language topic per the user constraint — none present. 9. Wrote the two-section digest to /scratchpad/digest/d12-dict.md in encounter order.

**Anomalies:** No true garbled/unreadable stretches. (1) File begins cleanly at OUTBOARD (adj), a fresh headword with full definition — NOT a mid-entry continuation, so no entry was skipped at the top boundary. (2) plug (v): the SEAL (v) alternative's technical-name parenthetical was column-interleaved in the raw text as "(WITH A PLUG [TN] OR A PLUGS [TN])"; reconstructed as "(WITH A PLUG [TN] OR PLUGS [TN])" — the stray "A" before PLUGS is a pdftotext layout artifact. (3) preferable (adj): the STE/Non-STE example prose was lightly garbled by the pdf ("...same type. BUT THE..." / "...but. according to the Standard Practices Manual. it is always preferable..."), but the approved alternative RECOMMEND (v) is unambiguous and unaffected. (4) Multi-meaning approved headwords PART (n), POINT (n), POINT (v) each occupy two lines in the digest (one per numbered sense) but are counted as one entry each. (5) PERFORMANCE (n) is approved for its given meaning; its embedded "For other meanings, use: DO (v)" guidance is captured as a bracketed note on the approved line, not as a separate entry. No inclusive-language topic found anywhere in this batch, so the skip line was not needed.

## d13-dict.md - Part 2 Dictionary word list, pdf pages 341-356 (page markers 2-1-P13 through 2-1-R8, with a Blank Page at 2-1-Q2). Headw

1. Read the entire source file (758 lines) top to bottom, covering page markers 2-1-P13 (PROCEDURE) through 2-1-R8 (RELATION). 2. Ran wc -l to confirm 758 lines = whole file, no unread tail. 3. Confirmed the file opens with a complete headword (PROCEDURE (n) = "Steps that are in a sequence"), so no continuation-from-previous-batch entry needed skipping. 4. Extracted letter-P tail entries P13-P18: approved PROCEDURE, PROPERTY, PROPORTION, PROTECTION, PROTECTIVE, PROTRUDE, PULL, PUSH (2 senses), PUT, PUT ON; not-approved proceed, process (prep/n/v), produce, product, profile, program, progress (n/v), progressive, progressively, prohibit, prolonged, prompt, propagation, proper, properly, proportional, proportionally, protect, provide, provided(that), providing(that), proximity, pump, puncture, purge, purify — preserving TN markers (WITH A PUMP [TN]) and cross-refs. 5. Extracted letter-Q entries P18/Q1 (Blank Page at Q2): approved QUALITY, QUANTITY, QUICKLY; not-approved qualified, quarterly, questionable, quick, quit, quote. 6. Extracted letter-R entries R1-R8: approved RADIAL, RADIALLY, RADIOACTIVE, RANDOM, RANDOMLY, RANGE, RATE, READ, REAR (adj + n = 2 entries), REARWARD, RECEIVE, RECESS, RECOMMEND, RECORD (2 senses), RECYCLE, REFER (2 senses), REFLECTION, REFUEL, REGULAR, REGULARLY, REJECT, RELATED, RELATION; not-approved raise, rapid, rapidly, re-(prefix), reach, reactivate, reading, readjust, ready (adj/v), real, rear of, rearmost, reason, reassemble, rebuild, recess, recharge, recheck, reclaim, recleat, reclose, recoil, recondition, reconstruction, recover, rectify, recur, reduce, reference, referenced, refill, refit, reflect, register, regrease, regulate, regulation, reinflate, reinforce, reinstall, reinstallation, rejection, relate. 7. Classified each headword by case (UPPERCASE = APPROVED with its column-2 definition; lowercase = NOT APPROVED with its uppercase alternatives), reassembling wrapped column text and compressing rewording-style guidance into [square-bracket] hints (product, re-, process TN note, progress IN PROGRESS ref, propagation limits note). 8. Wrote the two-section digest in encounter order and verified counts with awk/grep: 37 unique approved headwords (40 meaning lines), 79 not-approved, sum 116.

**Anomalies:** No inclusive-language material in this batch (grep confirmed), so nothing skipped on that account. File starts cleanly at PROCEDURE (n) with a full definition — NOT a mid-entry continuation, so no leading entry was dropped. A "Blank Page" marker sits at line 352 (page 2-1-Q2) between the Q and R sections; ignored as instructed. Minor pdftotext typos exist ONLY in the non-extracted STE EXAMPLE columns and do not affect any meaning/alternative: line 189 "FO THE COMPONENT" (for "FOR"); lines 231-236 "REMOVE THE OIL FROM THE TANK AN APPLICABLE PUMP ... CIRCUIT LINE" (appears to drop "WITH"/"WITH AN"). All headword meanings and approved alternatives reconstructed with full confidence; no garbled column text in the extracted fields.

## d14-dict.md - Part 2 word list, headwords "relay" through "separable" (tail of R section + head of S section). PDF pages 357-372 = dic

1. Read the entire source file (745 lines) top to bottom, page 2-1-R9 through 2-1-S6, stripping repeated headers/footers ("ASD-STE100...", "Issue 9", "2025-01-15", "Page 2-1-Rx/Sx") and the "Word / (part of speech) / Approved meaning/ALTERNATIVES / STE EXAMPLE / Non-STE example" column-header rows. 2. Worked the tail of the R section (pages 2-1-R9..R18): relay through running - classified each headword by case (UPPERCASE=approved, lowercase=not approved) and reassembled wrapped column-2 text; approved here: RELEASE, REMAINING, REMOVAL, REMOVE, REPAIR (n & v), REPLACE, REPLACEMENT, REPORT, RESISTANT, RESULT, RETRACT (2 meanings), RETRACTION, RIGHT, RIGID, RISK, ROUGH, ROUNDED, ROUTING, ROW, RUB. 3. Confirmed the first line "relay (v)" is a complete entry (SEND alternative + both examples present), so there was no orphaned continuation from the previous batch to skip. 4. Handled multi-alternative not-approved R entries carefully (e.g. restrain -> HOLD/STOP/PREVENT/LIMIT; restrict -> DECREASE/PREVENT/ONLY/LIMIT; reposition -> MOVE/PUT/SET) and note-type alternatives (reverse, rework). 5. Ignored the "Blank Page" marker at source line 464 (page 2-1-R18). 6. Worked the head of the S section (pages 2-1-S1..S6): SAFE through separable - approved: SAFE, SAFELY, SAFETY (n & v, v has 2 meanings), SAME (adj & pron), SAMPLE, SATISFACTORILY, SATISFACTORY, SCHEDULE, SEAL (n & v), SECONDARY, SEE, SELECT, SELECTION, SEMICIRCULAR, SEND, SENSE, SENSITIVE; captured the (TN)-qualified alternatives for safety-clip/safety-wire/scrape/screw/scrub and the "For other meanings use" list under SEE plus the "not a synonym for SET" note under SELECT. 7. Scanned the whole batch for any inclusive-language material - found none - so no skip line was written. 8. Split into APPROVED (41) and NOT APPROVED (92), preserving original order, one line per meaning/alternative set, and wrote the digest; counted the two sections separately (41 + 92 = 133).

**Anomalies:** No inclusive-language material in this batch (nothing skipped for that reason). No garbled/unreconstructable column text; the 4-column layout was clean and every entry reassembled with confidence. The batch starts cleanly at headword "relay (v)" (line 5) with a full definition, so NO mid-entry continuation was inherited from the previous batch and nothing was skipped at the boundary. The batch ends cleanly at a complete entry "separable (adj)" (last full page 2-1-S6). Notable multi-part entries preserved: RETRACT (v) and SAFETY (v) each carry two numbered meanings (repeated on separate lines); REPAIR, SAME, SEAL each appear under two parts of speech. SEE (v) and SELECT (v) are APPROVED but carry usage restrictions, captured inline in square brackets. Two entries had a note (not a word) as their alternative, compressed to [square brackets]: reverse (adj) [part of a technical noun for rearward thrust] and rework (v) [refer to technical verbs; manufacturing processes]. Technical-name/verb markers (TN) preserved on alternatives: ROPE, SCAN, SCORE, SCRATCH, plus parenthetical TN qualifiers on safety-clip, safety-wire, scrape, screw, scrub. One "Blank Page" marker was ignored at dictionary page 2-1-R18 (source line 464). Approved-verb inflection lists (e.g. RELEASES/RELEASED) and adjective comparatives (ROUGHER/ROUGHEST) were retained in [brackets] for fidelity.

## d15-dict.md - Part 2 Dictionary word list, letter S: from "separate (adj)" through "station (v)" (pdf pages 373-388, book pages 2-1-S7

1. Read the entire source file d15-dict-p373-388.txt (801 lines) top to bottom; ran wc -l to confirm 801 lines = full batch, no truncation. 2. Confirmed the leftmost-column headword on line 5 ("separate (adj)") is a fresh, complete entry (headword + pos + approved meaning + full example), not a continuation with no headword, so nothing was skipped at the top boundary. 3. Grepped for "inclusive" to confirm no inclusive-language topic exists in this batch (none), so the skip-line rule did not apply. 4. Walked every leftmost-column headword in encounter order across the S range: separate/separation/sequence; serious/serrated/serration/serve/service/serviceability/serviceable/servicing; set/setting/settle/sever/several/severe; shake/shall/shape/sharp/sharply/shear/sheet/shield/shift/shiny/shock/short/shorten/should/show/shut down; side/sideways/sight/sign/significant/silence/similar/simulate/simultaneous(ly)/since; situated/situation/size; skid; slack/slacken/slackness/slave/slide/slight(ly)/slip/slope/slot/slow/slow down/slowly/sluggish; small/smear/smell/smoke/smooth(ly); snag/snap; so (that); soak/soapy/soft/soften/soiled/solid/solidify/solution/solve; some/something/soon/sound/source; space/spacing/spanwise/sparingly/spark; speak/special(ly)/specific(ally)/specified/speck/speech/speed/spherical/spill/spillage; spin/splash/splined/split/spot/spray/spread/spring/spurious; stability/stabilization/stabilize/stable/stage/stained/stamp/stand/standard/starboard/start/state/station. 5. Classified each by headword case: UPPERCASE -> APPROVED section (transcribed column-2 definition, terse; repeated headword on its own line for each numbered sense of SHOW, SIDE, SMOOTH, SOAK, SOLUTION, SOURCE, START; carried inflection forms and comparative/superlative forms). 6. lowercase -> NOT APPROVED section (transcribed every column-2 ALTERNATIVE in order, preserving (pos), TN markers, uppercase multi-word phrasings; compressed pure rewording instructions to [square-bracket] hints for serve, several, situation, soon). 7. Counted 54 approved + 89 not-approved = 143 and wrote both sections to the digest.

**Anomalies:** No garbled/unreconstructable text - all 4-column entries reassembled cleanly despite column interleaving. No inclusive-language material present in this batch (nothing skipped per that constraint). First entry "separate (adj)" (line 5) is a complete headword, NOT a mid-entry continuation from the previous batch, so no leading entry was skipped. Last entry "station (v)" (lines 794-796) is complete, ending mid-alphabet at "sta-" (next batch presumably continues with "stay"/"steady" etc.). Notes captured inline: SHORT carries "[if possible, be accurate]"; SINCE carries "for other meanings use BECAUSE (conj)"; several/situation/soon/stamp carry their usage-note guidance in [brackets] or parens. Uppercase multi-word alternatives (NOT CONNECTED, NOT ATTACHED, ALMOST THE SAME, AT THE SAME TIME, CAUSE TO FALL, MOVE ACCIDENTALLY, MOVE QUICKLY, TURN...QUICKLY, MAKE...SOFT) preserved verbatim as approved-word phrasings. Technical names present as alternatives: SKID (TN), TENSION (TN), SOAP (TN), SPAN (TN), SPLINE (TN), STAIN (TN), STAMP/STAMPS (TN).

## d16-dict.md - Part 2 Dictionary word list, headwords "stationary" through "till" (letters S-T). PDF pages 389-404 = dictionary pages 2

1. Read the entire source file /tmp/.../spec/d16-dict-p389-404.txt (785 lines) top to bottom in a single pass. 2. Identified the 4-column table layout (Word/POS | Approved meaning-ALTERNATIVES | STE EXAMPLE | Non-STE example) and stripped repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Page 2-1-Sxx/Txx"). 3. Confirmed the batch opens on a complete headword (stationary, S23) so nothing carries over mid-entry from the prior batch. 4. Walked every leftmost-column headword page by page: S23 stationary->stiff; S24 still->STOW; S25 STRAIGHT->stroke; S26 STRONG->SUBSEQUENT; S27 SUBSEQUENTLY->SUBTRACT; S28 successful->suitable; S29 suitably->SURFACE; S30 surplus->sweep; S31 swing->synchronize; S32 SYNCHRONIZED, SYSTEM; T1 TACKY->TAP; T2 tape->tension; T3 terminate->THEN; T4 THERE->THICKNESS; T5 THIN->THROUGH(prep); T6 throughout->till. 5. Classified each by headword case: UPPERCASE -> ## APPROVED with column-2 meaning; lowercase -> ## NOT APPROVED with column-2 alternatives (kept UPPERCASE, POS, TN/TV markers exactly; compressed the two parenthetical rewrites (WITH STRAPS [TN]), (WITH SUCTION [TN]) as given). 6. Reassembled wrapped meaning text across lines (e.g. TEST n, SYMBOL n, SUBASSEMBLY n) and preserved multi-meaning rows as separate "=" lines. 7. Preserved usage-redirect and cross-reference notes in [brackets] on the affected approved entries. 8. Checked for inclusive-language content per the user constraint - none present in the S-T range. 9. Counted 68 approved + 65 not-approved = 133 entries and wrote the two-section digest in encounter order.

**Anomalies:** No garbled or unreconstructable text - all 4-column rows reassembled cleanly. No inclusive-language material in this batch (nothing skipped for that reason). The file begins with the complete headword "stationary (adj)" (page S23) - NOT mid-entry - so no continuation entry from a previous batch was skipped. Note on counting: entries are counted per headword+part-of-speech row (each dictionary row = 1 entry); headwords appearing under multiple parts of speech are counted as separate entries (e.g. SUPPLY n + SUPPLY v = 2; THAT conj + THAT pron = 2; THERE adv + pron; THESE adj + pron; THIS adj + pron; THROUGH adv + prep; substitute adj/n/v = 3; support n/v; suspect adj/v). Multiple numbered meanings within a single row (STOP, STRUCTURE, SUPPLY v, SYNCHRONIZED) are one entry but transcribed as separate "= " lines. Four approved words carry usage-redirect notes that I preserved in [brackets] rather than dropping: SUBSEQUENT (use NEXT), SUBSEQUENTLY (use THEN), THUS (use FOLLOW for other meanings), TIGHT (use SEAL/TENSION for other meanings); also SWALLOW (safety instructions only), and GR cross-refs on THAT-conj (GR-1) and THIS-pron (GR-4). "such as" has no part-of-speech marker in the source (transcribed as-is).

## d17-dict.md - Part 2 Dictionary word list, headwords T (from TILT) through U (to utilize), pdf pages 405-420 / dictionary pages 2-1-T7

1. Read the entire source file /tmp/.../spec/d17-dict-p405-420.txt (lines 1-751) top to bottom, ignoring repeated page headers/footers ("ASD-STE100 Simplified Technical English", "Issue 9", "2025-01-15", "Page 2-1-Tx/Ux"), column-header rows, and the "Blank Page" marker at 2-1-T14. 2. Confirmed the batch opens on a complete headword (TILT (v)) rather than continuation text, so no prior-batch entry needed skipping. 3. Reassembled each wrapped 4-column entry (Word/pos | Approved meaning or ALTERNATIVES | STE EXAMPLE | Non-STE example) into single logical entries across the T range: TILT, TIME, time, tip(n), tip(v), TO, TOGETHER, TOLERANCE, TOO, TOOL, TOP(adj/n), top off, TORQUE, TOTAL(adj/n), TOUCH, TOW, toward, toxic, trace(n/v), track, trained, transfer(n/v), transition, TRANSMIT, TRANSPARENT, transport, trapped, TRAVEL, travel, treat, trigger, trim, trip, troubleshoot, true, TRY, tuck, TUNE, TURN(n/v), turn off/on/over, twice, TWIST, TYPE, TYPICAL. 4. Continued through the U range: unable, unauthorized, uncap, unclip, uncoil, uncontaminated, uncouple, uncovered, undamaged, UNDEMANDED, under, underneath, underside, undertake, undo, undue, uneven, UNFOLD, uniform, uniformly, unintentional, unintentionally, UNIT, UNKNOWN, unlatch, UNLESS, unload, UNLOCK, unnecessary, unobstructed, unpack, unpainted, unplug, unprotected, unreel, unrestricted, unsafe, UNSATISFACTORILY, UNSATISFACTORY, unscrew, UNSERVICEABLE, unstable, UNTIL(conj/prep), unused, UNUSUAL, UNUSUALLY, UNWANTED, UNWIND, UP(adj/adv/prep), upon, upper, uppermost(adj/adv), UPSTREAM(adj/adv), UPSTREAM OF, up to, upward, urgently, usage, use(n), USE(v), USUAL, USUALLY, utilization, utilize. 5. Classified each by headword case: UPPERCASE -> APPROVED section with its terse column-2 meaning; lowercase -> NOT APPROVED section with its approved alternative(s) preserving (v)/(n)/(adj)/(adv)/(TN) markers exactly, joined by semicolons. 6. Checked the batch for any inclusive-language material per the user constraint; none present in this T-U word-list range, so nothing was skipped on that basis. 7. Wrote the digest to the out path in the required two-section format and verified counts with grep: 48 distinct approved headword+pos entries (51 meaning lines) and 72 not-approved entries, summing to items=120.

**Anomalies:** Clean batch, no garbled text requiring reconstruction. The file begins mid-page-table but at a full headword (TILT), so there is NO continuation entry carried over from the previous batch to skip. A "Blank Page" marker at dictionary page 2-1-T14 (between the end of the T section and the start of the U section) was ignored as instructed. Three approved headwords carry two meanings each, recorded as separate meaning lines under one entry: TIME (n) [duration / occurrence], TWIST (v) [force causing distortion / turn or change shape by torsion], UNIT (n) [equipment doing a function / standard of measurement]. A few not-approved alternatives are phrases rather than single approved words and were kept verbatim as printed: unauthorized (adj) -> NOT APPROVED; undamaged (adj) -> NOT DAMAGED; unnecessary (adj) -> NOT NECESSARY; unstable (adj) -> NOT STABLE; under (prep) -> ...LESS THAN; up to (prep) -> ...; trip (v) -> ...CAUSE TO FALL. Parenthetical usage guidance preserved: uncap (v) -> REMOVE (v) (A CAP (TN) OR CAPS (TN)); unclip (v) -> REMOVE (v) (FROM A CLIP (TN) OR CLIPS (TN)). undertake (v) carries an inline note "You can use a more accurate verb" attached to its DO (v) alternative, recorded as [use a more accurate verb]. TOGETHER (adv) is approved with meaning "In one group or position" plus a printed note that for other meanings AT THE SAME TIME should be used, captured in square brackets.

## d18-dict.md - Part 2 word list, headwords V-Z (valid through zero), pdf pages 421-434. All 89 headword rows transcribed: 37 approved (

1. Read the whole source file /tmp/claude-1000/.../spec/d18-dict-p421-434.txt top to bottom (lines 1-599); confirmed it is the tail of the Part 2 Dictionary word list running V-Z. 2. Grepped for inclusive-language keywords (inclusive/gender/he-she/his or her/s-he) — zero matches — so the user's skip constraint did not trigger and all content was extracted. 3. Verified the batch opens on a fresh headword ("valid (adj)", line 5) rather than mid-entry, so no leading continuation needed skipping. 4. Worked the V range (valid, validity, VALUE, variable, various, vary, vent, ventilate, ventilation, verify, VERTICAL, VERTICALLY, VERY, via, vibrate, VIBRATION, vicinity, VIEW, view, vigorous, vigorously, visible, VISUAL, VISUALLY, vital, VOICE, volatile, VOLUME), classifying each by case (UPPERCASE=approved / lowercase=not) and reassembling the wrapped meaning/ALTERNATIVES column. 5. Worked the W range (WAIT, WALK, WANT, WARM, warm, warn, wash, watch, watertight, way, WE, WEAK, weakness, WEAR, WEATHER, weep, WEIGH, WEIGHT, well, WET, WHEN, whenever, WHERE, whether, WHICH, whichever x2, WHILE, whilst, WHO, whoever, whole, wholly, whose, WIDE, WIDTH, WILL, WIND, windy, wipe, wire, wire-lock, WITH, withdraw, within, WITHOUT, withstand, WORK, work, would, wrap, WRITE, wrong, wrongly), capturing irregular verb forms (WORE/WORN, WROTE/WRITTEN, WOUND) and comparatives (WEAKER/WEAKEST, WETTER/WETTEST, WIDER/WIDEST), and the TN/TV markers (FORCE (TN), WARNING (TN), WIND (TN), WITH WIRE/LOCKWIRE/SAFETY WIRE (TN)). 6. Worked the Y-Z range (YES, yet conj, yet adv, YOU, YOUR, yourself, zero), across two intervening blank pages. 7. Compressed the four rewording-only alternatives (volatile, whose -> use a different construction; and the note text on WE/WILL/WEAR restrictions) into square-bracket hints per format spec. 8. Wrote two sections (APPROVED, NOT APPROVED) in encounter order, one line per meaning/alternative set, then counted 37 approved headwords and 52 not-approved headwords, items=89.

**Anomalies:** No garbled text; the pdftotext -layout column wrapping was clean and every entry reassembled confidently. First entry "valid (adj)" starts fresh at the top of the batch (not a mid-entry continuation), so nothing was skipped as belonging to a prior batch. No inclusive-language content exists anywhere in this V-Z batch, so the skip rule did not apply. Notes on multi-meaning entries: VOLUME (n) has two numbered meanings (1. space an object fills; 2. loudness) and WEIGH (v) has two numbered meanings (1. measure weight; 2. have a specified weight) — each is rendered as two lines under one headword. "whichever" appears as two separate rows (adj and pron), both -> THAT (pron); "yet" appears as two separate rows (conj -> BUT; adv -> AT THIS TIME) — all four counted as distinct entries. Two "Blank Page" markers (pdf pages after W8 and after Y1) and repeated page headers/footers were ignored as instructed. Entry counting note: the 37 approved headwords span 39 digest lines because VOLUME and WEIGH each add a second meaning line.