Subsection Polisher

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass, no license.

.claude/skills

Audit 0.4.1 · source sha 81994516a430 · 19 Oct 2026Report an issue or request a re-scan

• PublisherWILLOSCAR GitHub userOwner of the repository this skill is read from
• Source folderWILLOSCAR/research-units-pipeline-skills/.codex/skills/subsection-polisherThe exact folder we audited
• LicenceNot declaredAs the repository states it
• Commit pinned for this verdict81994516a430A new audit runs when the source changes
• Repository statusNot archivedLast push 14 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Polish a single H3 unit file under `sections/` into survey-grade prose (de-template + contrast/eval/limitation), without changing citation keys.

---
name: subsection-polisher
binding: library
description: "Polish a single H3 unit file under `sections/` into survey-grade prose (de-template + contrast/eval/limitation), without changing citation keys."
---

# Subsection Polisher (local, pre-merge)

## Triggers & routing

- **Trigger**: subsection polisher, per-subsection polish, polish section file, 小节润色, 去模板, 结构化段落.
- **Use when**: `sections/S*.md` exists but reads rigid/template-y; you want to fix quality locally before `section-merger`.


Purpose: upgrade one `sections/S<sub_id>.md` (H3 body-only) so it reads like survey prose **before** you merge into `output/DRAFT.md`.

This is intentionally local: fix one unit at a time, rerun gates, and converge without rewriting the whole paper.


## Role cards (use explicitly)

### Local Section Editor

Mission: improve one H3’s argument density and paper voice without changing citation keys.

Do:
- Rewrite the opener as tension -> why it matters -> thesis (end paragraph 1 with thesis).
- Add explicit contrasts and one evaluation anchor when missing.
- Add a subsection-specific limitation that changes interpretation.

Avoid:
- Adding/removing citation keys or moving citations across subsections.
- Replacing content with generic boilerplate.

### Evidence Steward (stop padding)

Mission: prevent polishing from turning into invention when evidence is thin.

Do:
- If you cannot write a contrast or evaluation anchor without guessing, stop and route upstream.

Avoid:
- Strengthening claims beyond what the existing citations can support.


## Role prompt: Local Section Editor (one H3 at a time)

```text
You are editing one survey subsection to make it read like paper prose.

Your goal is to remove generator voice and strengthen argument moves without changing citation keys:
- opener: tension + why-it-matters + thesis (no narration)
- add explicit contrasts and an evaluation anchor if missing
- make at least one cross-paper synthesis paragraph (>=2 citations)
- add a subsection-specific limitation (not boilerplate)

Constraints:
- do not add/remove citation keys
- do not invent facts
- keep scope local to this H3
```

## Inputs

- Target file: `sections/S<sub_id>.md` (H3 body-only)
- Preferred context: `outline/writer_context_packs.jsonl`
- Fallback context: `outline/subsection_briefs.jsonl` + `outline/evidence_drafts.jsonl`
- `citations/ref.bib`

## Output

- Updated `sections/S<sub_id>.md` (same path; citation keys unchanged)

## Non-negotiables (contract)

- Citation keys are immutable: do not add/remove any `[@BibKey]` markers.
- Scope is immutable: keep all citations within this H3’s allowed scope (`outline/evidence_bindings.jsonl` / writer pack `allowed_bibkeys_*`).
- No invented facts: if you cannot write a concrete contrast or evaluation anchor without guessing, stop and fix upstream evidence.
- Body-only: no headings; `section-merger` adds headings.

## Target quality (what “polished” means)

A polished H3 reads like an argument, not a topic list:
- Paragraph 1 ends with a **thesis** (conclusion-first takeaway) and does not use narration templates.
- At least **two explicit contrasts** (A vs B) using contrast words.
- At least **one evaluation anchor** paragraph (task/benchmark + metric + constraint/budget/tool access when relevant).
- At least **one cross-paper synthesis** paragraph with >=2 citations in the same paragraph.
- At least **one limitation/caveat** tied to protocol mismatch / missing details / unclear threat model (not boilerplate).

## Paper voice constraints (high signal anti-patterns)

Delete / rewrite these (they read like a generator):
- Outline narration: `This subsection ...`, `In this subsection, we ...`.
- Slide navigation: `Next, we move ...`, `We now turn to ...`, `In the next section ...`.
- Meta guidance: `survey synthesis/comparisons should ...`.
- Evidence-policy disclaimer spam: repeated `abstract-only/title-only/provisional` boilerplate inside H3.
- Count-based slot openers: repeated "Two limitations..." / "Three takeaways..." used as paragraph starters.

Prefer these (paper voice):
- Content-first openers: `A central tension is ...`, `In practice, ...`, `One recurring pattern is ...`.
- Argument bridges (not navigation): `This contrast matters because ...`, `These assumptions shape ...`.
- Embedded citations as evidence (no trailing dump tags).

## Workflow (one subsection)

1) Load the subsection contract
- Read this subsection’s pack in `outline/writer_context_packs.jsonl`.
- If the pack is missing/thin, fall back to `outline/subsection_briefs.jsonl` (thesis/tension/paragraph_plan) + `outline/evidence_drafts.jsonl` (comparisons/eval/limitations).
- Extract (write down, not in the prose):
  - `tension_statement` + `thesis`
  - 2–3 `comparison_cards` you will use for A-vs-B contrasts
  - 1 `evaluation_anchor_minimal` (task/metric/constraint)
  - 1 limitation hook from the evidence pack

2) Preflight (kept out of the final prose)
- Draft 4 one-line sentences:
  - Tension
  - Contrast (A vs B; >=2 citations)
  - Evaluation anchor (task/metric/constraint)
  - Limitation
If you cannot write these without guessing, stop and push the gap upstream (`paper-notes` / `evidence-draft`).

3) Opener rewrite (paragraph 1)
- Remove narration openers.
- Write: 1–2 sentences tension/decision/lens + 1 sentence why it matters + end with the thesis.

Bad:
- `This subsection surveys tool interfaces for agents.`

Better:
- `A central tension in tool interfaces is balancing expressive action spaces with verifiable execution; interface contracts largely determine which evaluation claims are meaningful.`

4) Paragraph pass (argument moves > listing)
- Rewrite paragraph-by-paragraph using the `grad-paragraph` micro-structure:
- Best-of-2 rewrite (recommended): when a paragraph feels slot-like, draft 2 candidate rewrites and keep the one with clearer argument move + less template cadence (citations unchanged).
  - tension → contrast → evaluation anchor → limitation
- Ensure you includ

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/WILLOSCAR/research-units-pipeline-skills
git -C research-units-pipeline-skills checkout 81994516a430

Works in

1. 2026-10-19
   ✓SAFE B 89 / 100

   first audit · source 81994516a430

Questions about Subsection Polisher

• github.com/WILLOSCAR/research-units-pipeline-skills/tree/main/.c — the audited source at commit 81994516a430 and every finding aboveObserved · 19 Oct 2026

Spotted something wrong? Report a correction · Methodology