| name | verification-before-completion |
| description | Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims |
Verification Before Completion
Core principle: Evidence before claims, always.
The Gate
Before claiming any status (complete, fixed, passing, done):
- IDENTIFY — What command proves this claim? (test suite, linter, build, etc.)
- RUN — Execute the full command fresh in this session
- READ — Check full output and exit code
- VERIFY — Does output confirm the claim?
- NO: State actual status with evidence
- YES: State claim with evidence
- THEN — Make the claim, citing command + key output (e.g. "Ran
npm test: 34/34 pass, exit 0")
Skip any step = unverified claim. This includes satisfaction language ("Great!", "Done!", "Perfect!") and implications of success — not just explicit completion statements.
What Counts as Verification
| Claim | Requires | Not Sufficient |
|---|
| Tests pass | Test command output: 0 failures | Previous run, "should pass" |
| Linter clean | Linter output: 0 errors | Partial check, extrapolation |
| Build succeeds | Build command: exit 0 | Linter passing, "looks good" |
| Bug fixed | Reproduce original symptom: passes | Code changed, assumed fixed |
| Requirements met | Line-by-line checklist verified | Tests passing alone |
Red Flags — Stop and Verify
- Using "should", "probably", "seems to"
- Expressing satisfaction before running commands
- About to commit/push/PR without fresh output
- Trusting agent success reports without independent check
- Relying on partial verification
Regression Tests (TDD Red-Green)
Write test -> Run (pass) -> Revert fix -> Run (MUST FAIL) -> Restore -> Run (pass)
A regression test that was never seen to fail proves nothing.
Agent Delegation
When delegating to sub-agents: verify via VCS diff + running commands yourself. Agent success reports are claims, not evidence.
Bottom Line
Run the command. Read the output. Then claim the result. No shortcuts.
