engineering-polishing
Skill3 starsChanged 41 days agoMITView on GitHub
Polish, locally restructure, or translate source notes into English engineering manuscript prose while preserving evidence boundaries. Use when the user asks to improve English academic style, paragraph flow, claim strength, anti-AI wording, terminology consistency, hedging, clarity, or publication-ready expression for existing engineering conference or journal paper content. Do not use for creating a new section argument, figure/table-only work, reviewer responses, or readiness validation.
In this repository7 other files it gives its agents.
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 8 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Use this skill after the paper's claim and evidence are known.
- Distinctive: line 14
Do not invent facts, references, mechanisms, experiments, numbers, novelty, or limitations.
- Distinctive: line 16
Do not polish a broken argument as if it were correct.
- Distinctive: line 20
Prefer restrained, concrete academic English over promotional language.
- Distinctive: line 24
If the paragraph lacks a stable claim-evidence structure, first return a logic diagnosis and do not present polished prose as final.
- Very specific: line 26
Use engineering-writing when the user needs a new section argument, contribution map, or full paper structure.
- Very specific: line 28
Use engineering-figure-table for captions, figure/table roles, or visual consistency.
- Very specific: line 30
Use engineering-response for comments and response letters.
- Very specific: line 31
Use engineering-validation for build checks or readiness claims.
- Distinctive: line 71
Run a sentence-role check: each sentence must be necessary, placed correctly, connected to its neighbors, and bounded by evidence.
- Distinctive: line 73
Run a claim-strength diff against the source before accepting stronger verbs.
- Distinctive: line 111
If the user asks for output only, return only the polished prose unless doing so would hide unsupported claims or unverified evidence.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Engineering Polishing
Use this skill after the paper's claim and evidence are known. It improves English manuscript prose without hiding structural or evidence problems.
Core Stance
- Final output defaults to English manuscript prose.
- Do not invent facts, references, mechanisms, experiments, numbers, novelty, or limitations.
- Do not polish a broken argument as if it were correct.
- Diagnose the failure mode before rewriting.
- Preserve technical terms, variables, units, metric definitions, and evidence boundaries.
- Prefer restrained, concrete academic English over promotional language.
Boundaries
- If the paragraph lacks a stable claim-evidence structure, first return a logic diagnosis and do not present polished prose as final.
- Use
engineering-writingwhen the user needs a new section argument, contribution map, or full paper structure. - Use
engineering-figure-tablefor captions, figure/table roles, or visual consistency. - Use
engineering-responsefor comments and response letters. - Use
engineering-validationfor build checks or readiness claims.
When to Open Extra Files
| File | Open when |
| ../_shared/revision-moves.md | Revising a draft: order of revision, expert section moves, sentence-level rules |
| references/fact-boundary.md | The source is a rough skeleton, non-English notes, or evidence is incomplete |
| references/paragraph-function.md | Identifying paragraph job before rewriting |
| references/engineering-topic-modules.md | Polishing control, algorithm, data-driven, materials, process, or multi-physics paragraphs |
| references/list-handling.md | Skill-specific list cleanup examples |
| references/paragraph-flow.md | Paragraphs feel unclear, repetitive, poorly ordered, or hard to follow |
| references/claim-strength.md | Claims may overstate evidence, causality, robustness, generalization, or novelty |
| references/claim-strength-diff.md | Comparing source and rewrite to ensure polish did not strengthen claims |
| references/anti-ai-prose.md | Text sounds generic, repetitive, slogan-like, template-like, or model-generated |
| references/style-guardrails.md | Need sentence-level academic English, transitions, hedging, terminology, or mechanics |
| references/source-notes.md | Non-English or rough notes must become English manuscript prose |
| references/source-note-expansion-check.md | A rewrite may have introduced facts not present in source notes |
| references/whole-paper-style-drift.md | Multiple sections may sound inconsistent, overly smooth, or template-like |
| references/examples.md | Needing before/after examples or output behavior examples |
| references/failure-modes.md | Handling requests to strengthen unsupported claims or hide weak evidence |
| ../_shared/evidence-boundary.md | The rewrite may change factual scope or claim strength |
| ../_shared/claim-strength.md | A local claim needs verb or novelty calibration |
| ../_shared/ai-assisted-writing-policy.md | Polished prose may hide AI-smell, unverified facts, or disclosure risk |
| ../_shared/list-to-argument.md | Source prose is a list rather than a paragraph argument |
| ../_shared/non-english-source-notes.md | Source notes include assumptions or non-English shorthand |
| ../_shared/output-mode.md | The user asks for output only |
| ../_shared/sentence-role-and-story-flow.md | Checking whether each sentence is necessary, connected, and non-redundant |
| ../_shared/terminology-ledger.md | User asks for varied wording around technical terms |
Workflow
- Identify the section and paragraph job.
- Apply fact-boundary checks when source notes are thin.
- Diagnose the main problem: paper logic, section job, paragraph flow, claim/evidence mismatch, or sentence style.
- Fix higher-level logic before sentence polish: argument, then evidence design, then section moves, then sentences (revision-moves.md).
- Run a sentence-role check: each sentence must be necessary, placed correctly, connected to its neighbors, and bounded by evidence.
- Run a claim-strength diff against the source before accepting stronger verbs.
- Preserve source facts and evidence boundaries.
- Rewrite in English with concrete subjects, varied sentence shapes, and calibrated verbs. Put linking information at the start of a sentence and the new point at the end; keep subject and verb close. Check single words that change the technical claim ("dedicated", "verify", "generalize").
- Report any claim that still needs evidence.
A paragraph has a stable claim-evidence structure only when at least three of these are clear: paragraph job, main claim, evidence anchor, and boundary. If fewer than three are clear, diagnose first and do not present polished prose as final.
Default Output
Diagnosis
- Paragraph job:
- Main issue:
- Evidence risk:
Before / After / Rationale
| Source sentence or span | Revised sentence or span | Rationale | Fact/evidence risk |
Claim-strength diff
| Source claim | Revised claim | Strength change | Allowed? | Reason |
Sentence role audit
| Sentence/span | Job | Needed because | Connection to previous/next | Evidence boundary | Action |
Polished version
[English prose]
Unsupported claims
- ...If the user asks for output only, return only the polished prose unless doing so would hide unsupported claims or unverified evidence.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

