repair

name	repair
description	Apply explicitly approved safe repairs from a second-brain lint report and verify the result.
disable-model-invocation	true
allowed-tools	Read, Glob, Grep, Write, Edit, Bash(git diff *), Bash(git status *)

Repair Second Brain

Read the report or findings supplied in $ARGUMENTS. Apply only approved repairs.

Safe automatic repairs are exactly those listed under Change control in governance/policies.md. Require explicit human approval for every review-required change class in that policy.

Delegate changes to knowledge-integrator with operation type repair and an approved change package derived from the lint findings (include the intended knowledge/log.md entry in that package). Require knowledge-verifier (mutation mode) to pass on the full package. Do not mutate files after PASS. Retain unresolved findings in the report/summary.