# Common Writing Problems and Repair Map

This page turns recurring structural, expressive, and continuity problems in
long-form writing into executable audit modules.

## Design Principles

- Prioritize problems that recur across genres instead of hard-coding one sample.
- Treat phrases and sentence shapes as signals, not automatic deletion rules.
- Let genre, character voice, context, and intent decide whether wording should stay.
- Repair evidence, action, rhythm, and state change instead of merely swapping synonyms.

## Common Complaints

| Reader Complaint | Deeper Cause | Recommended Modules |
| --- | --- | --- |
| Too smooth, too complete, template-shaped | Paragraphs are symmetrical and every thought resolves cleanly | `formulaic-structure-audit`, `controlled-drift`, `vocal-rhythm` |
| Plastic stock phrases | Cliches replace observation, action, and proof | `cliche-phrase-audit`, `embodied-emotion`, `cultural-anchors` |
| Polished but hollow | Paragraphs restate the same premise without a new state change | `prose-progress-audit`, `editor-loop` |
| Big emotion, low realism | The draft names feelings without body, action, contradiction, or consequence | `embodied-emotion`, `relationship-state` |
| Dead transitions | Connective phrases replace causality, objects, action, or point of view | `narrative-bridges`, `formulaic-structure-audit` |
| Cultural vacuum | The prose lacks era, place, material detail, and community language | `cultural-anchors` |
| Period detail exists, but technology, institutions, and objects conflict | The setting is not treated as a testable constraint | `world-ontology-audit` |
| Investigation, research, negotiation, or combat is skipped | The result lacks attempts, resistance, judgment, evidence, and cost | `process-earnedness-audit` |
| Expansion grows while the promised process gets thinner | Attention is spent on atmosphere and semantic restatement | `attention-budget-audit` |
| Chapters change wording but repeat the same opening, turn, and hook | Chapter functions form a copied architecture | `chapter-pattern-audit` |
| Every scene ends with scenery, silence, reflection, or a lesson | The real ending has already occurred; a generic bookend simulates meaning without changing state | `earned-ending-audit`, deterministic `END001` |
| Long-form drift | Facts, relationships, space, props, and hooks are not tracked | `relationship-state`, `spatial-blocking`, `forensic-physical-audit` |
| False precision | Human perception is written like measurement data | `natural-measurement` |
| Repeated not-X/is-Y, is-X/not-Y, or chained Chinese 比 comparisons | Symmetry and escalation substitute for evidence or judgment | `formulaic-structure-audit` |
| Fluent-looking sentences omit a word, object, or connector clause | No final predicate-slot, parallel-structure, or reference pass | `proofreading-audit`, deterministic `lint` |

## How the New Modules Help

### Cliche Phrase Audit

`cliche-phrase-audit` checks for stock fate language, generic body cues,
decorative emotional labels, dead transitions, and common English AI-slop phrases.

The repair is not synonym swapping. Replace the weak phrase with a visible action,
specific object, local sensory fact, attributed evidence, character-specific verbal
habit, or causal bridge.

### Formulaic Structure Audit

`formulaic-structure-audit` checks for overused triplets, forward and reverse
contrasts, chained comparison criteria, identical paragraph cadence, and paragraphs
that close too neatly.

The repair is purposeful asymmetry: keep necessary order, vary paragraph function,
and let some pressure carry into the next beat. Necessary correction, legal exclusion,
sourced comparison, and character rebuttal remain valid. Ambiguous omissions move to
the final proofread pass instead of being guessed during style rewriting.

### Prose Progress Audit

`prose-progress-audit` asks what became newly true in every paragraph. If two
neighboring paragraphs reduce to the same summary, cut, merge, replace with
evidence, or convert explanation into action.

## Recommended Commands

Use `--deep-review` for the complete structure-and-expression audit set; use compact `--review` for long-form continuation:

```powershell
python -m humanwriting.cli build `
  --style fiction `
  --deep-review `
  --strict-continuity `
  --context examples/story-ledger.md `
  --task "Write the next scene. Avoid stock phrases and formulaic paragraph shapes."
```

Audit an existing draft:

```powershell
python -m humanwriting.cli audit `
  --draft examples/problem-car-scene-draft.md `
  --context examples/vehicle-scene-ledger.md `
  --profile full `
  --profile numbers
```

AI taste is not only a word-list problem. It emerges when structure, evidence,
rhythm, voice, and context maintenance fall out of balance.