eng-verification
1
总安装量
1
周安装量
#77489
全站排名
安装命令
npx skills add https://github.com/tjboudreaux/cc-plugin-engineering-excellence --skill eng-verification
Agent 安装分布
amp
1
cline
1
opencode
1
cursor
1
kimi-cli
1
codex
1
Skill 文档
Verification Before Completion
Overview
Claiming success without evidence is dishonesty, not efficiency. Every status report, âdone,â or âtests passâ statement requires fresh verification.
Iron Law
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
If you didnât run the verifying command in this message, you cannot claim it passes.
Gate Function
1. IDENTIFY: Which command proves the claim?
2. RUN: Execute the full command now (fresh run).
3. READ: Inspect entire output & exit code.
4. VERIFY: Does output prove the claim?
- If NO â report actual status with evidence
- If YES â state claim WITH evidence (command + summary)
5. ONLY THEN: Make the claim
Skipping any step = lying.
Common Claims & Required Evidence
| Claim | Required | Not Enough |
|---|---|---|
| âTests passâ | Latest test command output showing 0 failures | Previous run, âshould passâ |
| âLint cleanâ | Linter command exit 0 now | Partial subset, inference |
| âBuild worksâ | Build command exit 0 | Linter pass |
| âBug fixedâ | Reproduction test now passes | Code diff only |
| âRegression test existsâ | RedâGreen verified cycle | Test created but never failed |
| âAgent succeededâ | Inspect VCS diff + verify outputs | Agent success message |
Red Flags â STOP
- Saying âshould/probably/seemsâ
- Expressing satisfaction before running verification
- Planning to commit/push/PR without a fresh run
- Trusting agent reports without checking
- Partial verification (âjust ran affected testsâ)
- âJust this onceâ or âIâm tiredâ rationalizations
Rationalization Rebuttals
| Excuse | Reality |
|---|---|
| âShould work nowâ | Run the command |
| âIâm confidentâ | Confidence â evidence |
| âJust this onceâ | No exceptions |
| âAgent said successâ | Verify independently |
| âPartial check is fineâ | Partial proves nothing |
| âIâm exhaustedâ | Exhaustion isnât evidence |
Patterns
- Tests: run and show
X/Y passingoutput. - Regression tests: demonstrate fail (red), apply fix, pass (green), revert to confirm failure if necessary.
- Requirements: re-read spec, create checklist, tick items with evidence.
- Agent work: inspect actual diff, run verification yourself.
When to Apply
Always before:
- Claiming completion/success
- Expressing satisfaction (âGreat! Done!â)
- Moving to next task
- Committing, pushing, or filing PRs
- Delegating completion to others/agents
Bottom Line
Evidence precedes claims. Run the command, read the output, then speak. Anything else breaks trust.