Lean doc feedback
APM-installable agent skills for Lean 4 and Mathlib4 — proof tactics, math domains, review and research workflows, generic tooling.
npx -y skills add r-irbe/proof-skills --skill lean-doc-feedbackAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
USE FOR: reviewing a Lean documentation draft, flagging undefined terms, suggesting cross-links, scoring readability against a rubric. DO NOT USE FOR: rewriting prose end-to-end (use @lean-doc-improvement), generating new blueprints (use @lean-blueprint). TRIGGERS: doc-review, readability, doc-feedback.
SKILL.md
2.2 KB, as published. Nobody here has run it
lean-doc-feedback
Routing
- USE FOR: reviewing a Lean documentation draft; flagging undefined terms; suggesting cross-links to blueprint nodes; scoring readability against the project rubric.
- DO NOT USE FOR: rewriting prose end-to-end (use
@lean-doc-improvement); generating new blueprints (use@lean-blueprint). - TRIGGERS: doc-review, readability, doc-feedback.
Behavioural rules (G-*)
- G-1 (MUST): Feedback MUST be line-anchored to the draft. [Trace: AC-01]
- G-2 (SHOULD): Feedback SHOULD prefer minimal-diff suggestions over rewrites. [Trace: AC-02]
- G-3 (MUST NOT): The skill MUST NOT silently edit the draft. [Trace: AC-03]
- G-4 (SHOULD): The skill SHOULD link each finding to a rubric criterion. [Trace: AC-04]
Workflow
- Discover [discover] — read draft, locate rubric, list scope boundaries.
- Review [execute] — emit line-anchored findings.
- Score [validate] — apply rubric; max 3 attempts then escalate.
- Persist [persist] — write
feedback.md, ticktasks.md.
Recovery & STOP
- Draft missing rubric reference → STOP, escalate.
- Scope drift (silent edit) → immediate STOP, restore.
- Confidence < 80 % on rubric criterion → STOP, ask.
Handoffs
- Predecessors / successors: see FM
handoffs. - Source spec:
specs/lean/doc-feedback/requirements.md. - Related ADRs: ADR-0076.
Common failure modes
AI agents commonly: rewrite instead of annotating; omit rubric anchors; bundle many findings into one comment.