| name | ai-detect |
| description | Audit drafts, slides, reports, homework, and sendable external messages for AI-smelling wording using Mingdao's historical session feedback. Use when the user asks whether a deliverable has 'AI味', sounds unnatural, has high entropy, uses placeholder titles, over-explains, contains process filler, or does not sound like a normal human wrote it. |
AI Detect
Purpose
Audit drafts, slides, reports, homework, and sendable external messages for AI-smelling wording using Mingdao's historical session feedback. Use when the user asks whether a deliverable has 'AI味', sounds unnatural, has high entropy, uses placeholder titles, over-explains, contains process filler, or does not sound like a normal human wrote it.
Keep this SKILL.md as a concise routing and execution entrypoint. Do not load
long examples, command catalogs, detailed checklists, or edge-case policy until
the current task needs them.
Workflow
- Confirm the user request matches this skill's frontmatter description.
- Bind the concrete target: source file, artifact, repo, device, document,
dataset, or user-facing deliverable.
- Use the smallest relevant workflow from this entrypoint first.
- Before scanning a deliverable, applying historical rules, running bundled
scripts, rewriting high-risk wording, or reporting AI-smell status, read
references/entrypoint-details.md.
- Preserve local owner boundaries: route to a narrower skill or repo-specific
workflow when the detailed reference indicates a more specific owner.
- For dedicated redundancy diagnostics, route to the
redundancy subskill
under this skill instead of reimplementing low-information repetition checks.
Detailed Reference
Read references/entrypoint-details.md for:
- Overview
- Trigger Conditions
- Canonical Workflow
- Seeded Confirmed Rules
- User Evidence Base
- Bundled Scripts
- Quick Start
- Output Expectations
- Guardrails
- Validation And Checkpoints
Validation
- Use the skill-specific validation or acceptance checks from the detailed
reference before declaring completion.
- When editing this skill, run
quick_validate.py on the skill directory and
verify all referenced files still exist.