agentleFS
r-irber-irbe/proof-skills/skills/lean-mwe/SKILL.md

lean-mwe

Skill3 starsChanged 24 days agoApache-2.0View on GitHub

USE FOR: creating minimal working examples (MWEs) from Lean 4 errors for upstream bug reports. DO NOT USE FOR: bisecting which Lean version introduced the bug (@lean-bisect). TRIGGERS: MWE, minimal repro, "minimise the error", "#guard_msgs".

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

Files
SKILL.md

Minimizing Lean Errors

Agnostic Minimal Working Example (MWE) contract for Lean 4. Guides isolating tactic, elaborator, or compiler bugs into self-contained reproducers using #guard_msgs, #guard_panic, and LSP-guided dependency pruning.

Routing

  • USE FOR: wrapping a failing Lean fragment in a #guard_msgs / #guard_panic guard that captures the exact diagnostic, then driving lake exe minimize (or manual LSP-guided import pruning) to delete imports and definitions until a self-contained reproducer remains; verifying the result with lake env lean; resuming interrupted minimisations with --resume.
  • DO NOT USE FOR: bisecting which Lean commit broke the file (use @lean-bisect after the MWE is self-contained); building or running the live project

Workflow

  1. Set up the guard -- choose #guard_msgs (for errors / warnings) or #guard_panic (for panics). Paste the exact diagnostic text verbatim into the guard.
  2. Verify the guard -- lake env lean YourFile.lean must produce no output (guard passes). If it errors, the guard text does not match; adjust the guard before minimising. Always use lake env lean, never bare lean.
  3. LSP-guided import pruning -- before running heavy minimization, inspect the forward import tree via $/lean/moduleHierarchy/imports. Remove unused transitive imports to isolate the dependency closure.
  4. Minimise -- lake exe minimize YourFile.lean. Use --quiet for long runs; --resume after manual edits to .out.lean; --only-delete / --only-import-inlining for surgical passes.

Recovery & STOP

  • Guard fails on first verify -> STOP, the diagnostic text doesn't match the model's output character-for-character; widen the guard or fix the example before running the minimiser.
  • Minimiser hangs > 60 min with --resume not progressing -> STOP, hand to @lean-bisect if the failure may be version-dependent; otherwise reduce the input by hand.

Handoffs

  • Predecessors / successors: see FM handoffs. Typical inbound: @lean-proof (when a tactic fails unexpectedly) or @lean-build (when CI surfaces an unexpected compiler error). Typical outbound: @lean-bisect (find the introducing commit) or @lean-pr (file the issue/PR with the MWE attached).
  • Sister skill: @lean-bisect -- typically the next step after the MWE is Mathlib-free.

Detailed reference

Full encyclopedic reference, decision trees, and lookup tables live in references/lean-mwe-handbook.md .

See also

  • ../../references/lean-mwe-handbook.md -- Handbook Handbook
  • ../lean-gateway/SKILL.md -- Top-level skill router

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.