| name | authorial-voice-integrity |
| description | Use when a user asks to make research writing sound less AI-like, more human, more authorial, less generic, or asks about AI-writing, AIGC, detector scores, disclosure, de-AI, humanising, or lowering AI rate; improves academic voice and integrity without detector-evasion. |
Authorial Voice Integrity
Use this skill for academic or professional writing requests framed as:
- AI-looking prose, AI trace, AI-style wording, generic AI writing;
de-AI, humanise, make it less ChatGPT, lower AI rate, AIGC, AI detector;- requests to hide, soften, remove, or invent an AI-use disclosure.
Core Rule
Reframe detector-focused requests into academic voice, argument quality, and integrity.
Allowed:
- remove prompt residue, chatbot meta-text, placeholders, and generic AI-style phrasing;
- strengthen paragraph-level authorial judgement through clearer mini-claims, warrants, and evidence boundaries;
- improve the confirmed project register, such as UK academic style when relevant;
- preserve meaning, citations, facts, numbers, dates, participant details, official requirements, and source boundaries;
- explain what changed and why.
Not allowed:
- promise a lower AI-detection or AIGC-detection score;
- optimise for detector evasion, randomisation, synonym-swapping, stylistic noise, or deliberate imperfection;
- claim how detectors, markers, reviewers, readers, or platforms will judge authorship;
- hide unsupported claims behind smoother prose;
- hide, soften, remove, or invent AI-use disclosure statements;
- change source claims, references, ethics/compliance statements, or official requirements for style.
Workflow
For quick chat-level style help:
- State that the work will improve authorial voice and academic integrity, not detector scores.
- Rewrite for claim clarity, source boundaries, and project-appropriate style.
- Mention any evidence, citation, or disclosure issue that must go to another gate.
For formal academic or professional artifacts:
- Run source-first and Material Passport checks where the artifact makes formal claims.
- Run
academic-integrity-preflight.
- Run
cognitive-frameworks and academic-self-review-loop when argument quality matters.
- Run the authorial voice scanner when local tools are available:
python3 scripts/authorial_voice_scan.py --target path/to/draft.md
Use --strict if a detector-evasion or disclosure-risk request should block the workflow.
- Apply
uk-academic-writing-style or the confirmed project style.
- Apply
style-memory-and-revision-gate before delivery.
Scanner Boundary
scripts/authorial_voice_scan.py is not an AI detector. It flags wording and request-pattern risks:
- detector-evasion framing;
- disclosure hiding;
- chatbot residue;
- generic AI-style phrasing;
- inflated academic vocabulary;
- formulaic transitions;
- possible overclaiming.
A HOLD result means the workflow should be reframed or checked before delivery. It does not mean the writing is AI-generated.
Escalate To Other Gates
- Prompt residue, placeholders, fake references, unsupported claims, or misleading disclosure:
academic-integrity-preflight.
- Source support, citation metadata, or claim support:
dissertation-citation-audit.
- Paragraph quality, mini-claims, warrants, and progression:
academic-self-review-loop with research-wiki/WRITING_QUALITY_RUBRIC.md.
- Formal delivery:
material-passport, formal-delivery-guard, and document-quality gates.
