verification-before-completionlisted

# Verification Before Completion Do not claim success without fresh command evidence. ## Gate Before any completion claim: 1. Identify the command that proves the claim. 2. Run the full command now. 3. Inspect exit code and output. 4. State results exactly as observed. 5. **If the change includes a condition or gate that determines when something applies: explicitly state what it does NOT cover. If the answer reveals a gap that should be covered, fix it before proceeding.** ## Applies To - "Tests pass" - "Bug is fixed" - "Build succeeds" - "Ready to merge" - Any equivalent wording ## Not Acceptable - "Should pass" - "Looks good" - Trusting old outputs - Trusting subagent reports without verification ## Minimum Evidence Examples - Tests: command output with zero failures - Build: successful exit code - Bugfix: reproduction case now passes - Requirements: explicit checklist against plan ## Stub Scan (Implementation Tasks) When verifying completion of any task that created or modified production code, run a stub scan before claiming done: ```bash grep -rn "TODO\|FIXME\|placeholder\|NotImplementedError\|raise NotImplementedError" <src-dir> \ --include="*.ts" --include="*.js" --include="*.py" --include="*.go" --include="*.rs" \ | grep -v -i "test\|spec\|__tests__" ``` Adjust `<src-dir>` and `--include` patterns to the project's language and source structure. If any match falls in a file this task created or modified: the task is not done. Remove the stub or confi