---
name: editorial-review-guidelines
description: >
  Use this skill when performing editorial review of ENSDF comment lines.
  Checks grammatical precision, subject-verb agreement, hyphenation,
  technical terminology, and ENSDF text-notation correctness. Suitable for
  editorial checks, grammar reviews, manuscript audits, or ENSDF comment
  validation in Nuclear Data Sheets manuscripts.
---

# ENSDF Editorial Review Skill

Column/field rules: `.github/agents/ENSDF-Agent.agent.md`. Spot-check policy: `.github/copilot-instructions.md`.

**Scope:** `c`, `cL`, `cG`, `cB`, `cE`, `cN`, `cP`, `cQ` comment records and continuation comment lines. Skip data-record fields (`L`, `G`, `E`, `B`, `DP`).
**Action Policy:** Two options: (1) Check only. Report findings table. No edits or revisions. (2) Check and revise. Report findings table, then apply edits to comment lines.
- Flag letters, marks, and labels are human-controlled: parenthesized letters in an identifier (`cL J(A)$`, `cL E(B)$`, `cG E(B),RI(B)$`) name the col-77 flag of the records the comment applies to; such lines may be reworded, but the flag letter must never be changed.
- General and per-reference comments are editable, but keep edits surgical and meaning-preserving.

## Error Classes

### 1. ENSDF Notation

- **Superscript/subscript:** Symbol outside braces. Wrong: `{+13C}` → Correct: `{+13}C`. Detection regex (correct form): `{+[0-9]+}[A-Z][a-z]?`
- **Plain isotope tokens:** Wrong: `3He`, `36S(d,3He)` → Correct: `{+3}He`, `{+36}S(d,{+3}He)`. Scan: `(?<!\{\+)\b\d{1,3}[A-Z][a-z]?\b`; keep valid element symbols; exclude tokens followed by `|` (e.g., `2I|g`). **Mandatory:** after the regex scan, manually review every remaining candidate — regex misses tokens in dense prose.
- **Missing `{I}` on uncertainty:** Comment uncertainties require `{In}` or `{I+n-m}`. Three patterns:
  - Parentheses: `124(5) ps` → `124 ps {I5}`.
  - Bare `I` prefix: `|d=+0.05 I5` → `|d=+0.05 {I5}`.
  - Bare number (no `I`): `A{-2}=+0.05 5` → `A{-2}=+0.05 {I5}`.
  Scan for bare `I`: `(?<!\{)\bI\d{1,3}\b` — the `(?<!\{)` lookbehind excludes valid `{I3}` / `{I+7-5}` braces (otherwise all braced uncertainties false-positive). Scan for bare numbers: lone 1-3 digit integers after `=` or after a value with space, where context suggests uncertainty.
- **Unicode leakage:** No raw Unicode glyphs. Wrong: `μ`, `×`, `β`, `γ`, `θ`, `≈`, `≤`, `≥` → Correct: `|m`, `|*`, `|b`, `|g`, `|q`, `|?`, `|<`, `|>`. Example: `1500-μm-thick` → `1500-|mm-thick`.
- **Non-ENSDF unit spellings:** `ug` → `|mg` (microgram); `ug/cm2` → `|mg/cm{+2}` (microgram per square centimeter); `mg/cm2` → `mg/cm{+2}` (milligram per square centimeter), `cm3` → `cm{+3}` (cubic centimeter).
- **Chemical formula stoichiometry:** Integer coefficients must be subscripts. Wrong: `H3BO3`, `Sb2S3` → Correct: `H{-3}BO{-3}`, `Sb{-2}S{-3}`. Scan: `(?<!\{\+)\b[A-Z][a-z]?\d+\b`. Exclude 1:1 binaries (`NaCl`), multipolarities (`E2`, `M1`), and `B(E2)` quantities.
- **Mixed symbol-text:** Wrong: `γ-ray`, `β-delayed` → Correct: `|g-ray`, `|b-delayed`.
- **Proper-noun exceptions:** Skip well-known instrument names (e.g., `Gamma Array` as part of INGA) — not every `Gamma` match is a symbol-text error.
- **Leaked record tags:** Scan cols 10-80 for spurious ` cL `, ` cG `, ` L `, ` G ` (copy-paste artifact).
- **Unintended symbol prefixes:** `|resonance` renders as rho+esonance; verify intent.
- **Reaction ejectile comma:** `target(beam,ejectiles)` takes exactly one comma, separating beam from concatenated ejectiles. Wrong: `27Al(12C,|a,p|g)` → Correct: `{+27}Al({+12}C,|ap|g)` (ejectiles α+p+γ concatenated, no internal comma).
- **Inconsistent subscripts:** `A{-2}=0.5 A6=-0.1` → `A{-2}=0.5 A{-6}=-0.1`.
- **Mid-token line breaks:** `E{-p}(lab)` must not split across continuation lines.
- **Subscript used as negative exponent:** `10{-n}` (renders as subscript) must be `10{+-n}` (negative superscript). Scan: `10\{-\d+\}` . Applies to all scientific-notation contexts (`\|*10\{-`, `E\{-`, `×10\{-` in raw Unicode).

### 2. Grammar and Style

- **Capitalization:**
  - Top-block comments (before first data record, any type, (`c  `, `cL $`, `cG  $`, `cL J$`, `cL J(A)$`, etc.)) → uppercase.
  - Record-specific (after first data record comments) **with** field identifier (`cL E$`, `cL J$`, `cL T$`, `cG E$`, `cG RI$`, `cG M$`, etc.) → lowercase, unless first token is numeral, symbol, isotope token, or acronym.
  - Record-specific comments (after first data record comments) **without** field identifier (`cL $`, `cG $`) → uppercase (standalone statements).
  - **Exception:** `cP` and `cN` comments always uppercase.
  - Examples (top-block, before first L/G/E/B/DP record): `cL E$From a least-squares fit` ✓; `cG E,RI$From 2017Li03` ✓; `cB IB,LOGFT$From I(|g+ce)` ✓ — all uppercase.
  - Examples (record-specific, after first data record): `cL T$from lifetime measurement` ✓ (lowercase continuation); `cL J$from the Adopted Levels` ✓; `cL J$1+ from shell model` ✓ (numeral exception).
- **NUCID case on comment lines:** Cols 1–5 must be uppercase. Wrong: `34Si cL J$` → Correct: `34SI cL J$`.
- **Extra space after `$`:** `cL J$ from ...` → `cL J$from ...`. Scan: `\$\s`.
- **Dittography:** `\b(\w+)\s+\1\b` -- flag `the the`, `were were`, etc.
- **Subject-verb agreement:** NSR keys are singular. Wrong: `1972Hu10 report` → Correct: `1972Hu10 reports`.
- **Missing auxiliary verbs:** `spectra recorded` → `spectra were recorded`.
- **Adjective/noun errors:** `others levels` → `other levels`.

### 3. Punctuation and Lists

- Comma splice between independent clauses → semicolon.
- Oxford comma required in lists of 3 or more items; only one `and` (at final item).
- Wrong: `14.9 {I6} (R1), and 8.3 {I4} (R2), and 10.5 {I60} (R3)` → Correct: `14.9 {I6} (R1), 8.3 {I4} (R2), and 10.5 {I60} (R3)`.

### 4. Hyphenation

- Compound modifier before noun: `4-mm-long target`, `gamma-ray spectrum`, `R-matrix analysis`, `2-g/cm{+2} Pb target`.
- Predicative: no hyphen (`the target was 4 mm long`).
- Always: `half-life`, `L-transfer`, `L-transfers`. Noun: `gamma rays`; adjective: `gamma-ray energy`.
- Exception: keep `J=L transfer` unhyphenated when `J=L` denotes the `J=L`, `J=L-1`, or `J=L+1` analyzing-power cases; this is not the compound noun `L-transfer`.

### 5. Terminology and Spelling

**Facility and instrument names:**

| Wrong | Correct |
|-------|---------|
| `Van der Graaff`, `Van de Graaf`, `Van der Graaf`, `Van de Craaff` | `Van de Graaff` |
| `Cockroft-Walton` | `Cockcroft-Walton` |

**Nuclear-physics terms and spelling:**

| Wrong | Correct |
|-------|---------|
| `uising`, `usign` | `using` |
| `deexiting` | `deexciting` |
| `multiporities` | `multipolarities` |
| `grand-daughter` | `granddaughter` |
| `ohter` | `other` |
| `stoped` | `stopped` |
| `striped` | `stripped` |
| `·coeffcients·` | `coefficients` |

- `GXPF1A` — exact capitalization required (shell-model interaction name).
- **Chemical formulas:** Use `{-n}` subscripts for stoichiometry. `CD{-2}` (deuterated polyethylene) ≠ `Cd{-2}` (cadmium); verify against publication.

### 6. Text and Number Integrity

- **Extra space after `=`:** `|w|g= 0.45` → `|w|g=0.45`. Scan: `=\s[0-9]`.
- **Missing space surrounding `=`:** `Intensity=0.45` → `Intensity = 0.45`.
- **Space within number:** `E{-p}(lab)=54 6` → `E{-p}(lab)=546`. Note: space between value and uncertainty is correct.
- **Field cross-contamination:** Energy in `RI$` or intensity in `E$` is an error. Wrong: `cG RI$ weighted average of 1224.6 {I154} ...` (contains energy, not intensity).

### 7. Logical Clarity

- Flag contradictory claims (confirmed and tentative in same sentence).
- Flag conclusions without citation or method reference.

## Cautionary Notes

Missing terminal periods, XREF notation. Valid ENSDF symbols: `|?`, `{+n}`, `{-n}`, `|a`, `|b`, `|g`, `|d`, `|w`, `|*`, `|+`, `|-`.
- Never strip a leading `|`: `|<` = ≤, `|>` = ≥, `|*` = ×. A leading `|` is never a stray artifact.
- Comment lines need not be padded to 80 columns (human wraps later); still confirm the diff is comment-only (col 7 = `c`).
- Comment-unit order inside one L/G block: identifiers run `E$` → `RI$` → `M$` → `MR$` → general (no identifier before `$`). A general comment sitting ahead of `M$`/`MR$` is a violation that `check_gamma_ordering.py` reports; read its COMMENT ORDERING section, not just the gamma/level ordering.
- Before reordering a checker-flagged comment block, verify its L/G-record boundary: comment-order findings can span adjacent levels and be false positives.
- Two datasets reporting the same value must share one clause; never demote an agreeing value to `Other:` (see the lifetime SKILL). Sentence-initial `Other:` is capitalized, identifier-adjacent `$other:` is lowercase.
- When a fix changes a line's trailing pad count, re-measure that line's length in the same turn: a replacement whose length differs from the matched span can relocate one space to the match boundary and yield 79/81-column lines.
- LaTeX/HTML leaked from a digest is a per-line repair: convert `$^{37}\text{Cl}(d,^3\text{He})^{36}\text{S}$` to `{+37}Cl(d,{+3}He){+36}S` and re-wrap the block (the replacement is shorter, so pads change).
- The 80-column all-space line that ends a file is part of the ENSDF submission format: never delete it, and treat a missing one as a repair candidate rather than noise.
- `d`/`D` (col 7) hidden-message records are free-text notes, not comments: never length-check or reword them.
- Values quoted in comments must name the specific gamma/level/multipolarity record and match it character-for-character — see `comment-quoted-values-check`.
- A capital that starts a new sentence (including after abbreviations such as `g.s.`, `m.s.`, `r.m.s.`) is not a capitalization error; do not lowercase it.
- Bare `|<`, `|>`, `|=` in comment should not convert to `<`, `>`, `=`.
- Column 78–79 findings are data errors, never `misplaced flags`: legal content there is L `MS` (`M `, `M1`/`M2`, `R `, `C `), G col-79 `N` (normalization), B/E `UN` (`1U`/`2U`), blank elsewhere. Never relocate these to column 77; `column_calibrate.py` and the ruler are record-type aware.
- G-record cols 33–76 are deliberately dense (M|MR|DMR|CC|DCC|TI all adjacent); `DCC` (63–64) touching `TI` (65–74) renders glued, e.g. `0.1656 24100` — do not insert a readability space or shift TI.

## Editing Discipline

- Use minimal single-line anchors; multi-line anchors spanning wrapped comment text can re-wrap lines and silently drop words.
- When two comment lines are byte-identical, disambiguate the anchor with the following line's text, and keep the padding arithmetic (insert chars / drop N pad spaces) inside the same edit. After editing, re-check the file's line count and every line's length, because a replacement that begins with a newline can lose it and merge two records.
- Comment-line layout: `NUCID` + continuation (col 6) + `c` (col 7) + record code (cols 8-9, e.g. `L `, `G `, or two blanks) + text at col 10, i.e. a 9-character prefix. A flag shifted into column 6 or column 8 makes the line unrecognizable (both directions are now reported by the validators), so assert `line[6]=='c'` and `line[9]!=' '` in any generating script instead of trusting a passing scan.
- After each file's edits, prove the diff is comment-only (changed lines have `c`/`d` in col 7) and no text was lost (word-level compare against `git show HEAD:<file>`).
- Find the top block by skipping lines whose col 7 is `c`/`d`; `P`/`N` records count as data records, and `cP`/`cN` stay uppercase.
- Add a recurring class to `scan_editorial_review.py` CHECKS with a `--selftest` case instead of writing ad-hoc sweep scripts; run it one path at a time.
- Report every edit as a literal before → after pair quoting the exact changed words, so the human user can track each change.

## Procedure

1. Run `python .github/scripts/scan_editorial_review.py [folder_or_file] --skip adopted` — automated sweeps (isotope tokens, bare `I`, braced `{I}`, units, chemical formulas, dittography, `$`/`=` space, `10{-n}`, unicode, leaked tags, spelling). Review every flagged line; discard false positives manually. The glob is one level deep: pass the folder that directly holds the `.ens` files, not its parent.
2. Apply error classes in order: notation → grammar → punctuation → hyphenation → terminology → integrity → clarity.
3. Manually scan the remaining lines for regex-missed issues (dense prose, capitalization, subject-verb, hyphenation) per the Error Classes above.
4. Final symbol sweep on all flagged lines: verify no raw Unicode glyphs, plain isotope tokens, mixed symbol-text compounds, or un-subscripted chemical formulas were missed.
5. Before reporting: confirm no valid ENSDF notation misclassified; each fix preserves scientific meaning. Do not edit unless user explicitly requests.

## Output Format

```
| File | Line(s) | Category | Current Text | Recommended | Rationale |
|------|---------|----------|--------------|-------------|-----------|
| f.ens | 47 | Subject-Verb | `predict` | `predicts` | NSR key is singular |
```