verification-before-completion
Agent skill fromobra/superpowers
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; evidence before assertions always
How to install verification-before-completion
install verification-before-completion
# install the superpowers pack with the skills CLI
npx skills add obra/superpowers# or add only this skill to Claude Code: copy the folder
skills/verification-before-completion/
# into
~/.claude/skills/verification-before-completion/ (all projects)
.claude/skills/verification-before-completion/ (one project)
1# Verification Before Completion23## Overview45**Core principle:** Evidence before claims, always.67**Violating the letter of this rule is violating the spirit of this rule.**89## The Iron Law1011```12NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE13```1415If you haven't run the verification command in this message, you cannot claim it passes.1617## The Gate Function1819```20BEFORE claiming any status or expressing satisfaction:21221. IDENTIFY: What command proves this claim?232. RUN: Execute the FULL command (fresh, complete)243. READ: Full output, check exit code, count failures254. VERIFY: Does output confirm the claim?26 - If NO: State actual status with evidence27 - If YES: State claim WITH evidence285. ONLY THEN: Make the claim2930Skip any step = lying, not verifying31```3233## Common Failures3435| Claim | Requires | Not Sufficient |36|-------|----------|----------------|37| Tests pass | Test command output: 0 failures | Previous run, "should pass" |38| Linter clean | Linter output: 0 errors | Partial check, extrapolation |39| Build succeeds | Build command: exit 0 | Linter passing, logs look good |40| Bug fixed | Test original symptom: passes | Code changed, assumed fixed |41| Regression test works | Red-green cycle verified | Test passes once |42| Agent completed | VCS diff shows changes | Agent reports "success" |43| Requirements met | Line-by-line checklist | Tests passing |4445## Red Flags - STOP4647- Using "should", "probably", "seems to"48…
Excerpt shown under the repository's MIT license.
