verification-before-completion
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
- License 许可
- MIT
- 最近更新
- 2026-08-20
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- verification-before-completion
- tags
- [ "workflow", "code-quality" ]
- tools
- []
- version
- 1.0.0
- agent_types
- [ "planner", "worker", "reviewer" ]
- 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; evidence before assertions always
- dependencies
- []
03 SKILL.md 正文
Verification Before Completion
The Iron Law
<EXTREMELY-IMPORTANT> No completion claim without evidence. "Done" means the named verification command ran, exited 0, and its output was inspected. Not "should work", "looks right", or "tested locally". Evidence before assertion, always. </EXTREMELY-IMPORTANT>
Use
Apply before any "done", "fixed", "passing", "works", or "ready to merge" claim; commit, push, PR, or after non-trivial edits. For pure prose or a directly observable artifact, cite the diff or artifact instead.
Evidence
| Claim | Required evidence |
|---|---|
| Test, typecheck, lint, or build passes | Named command, exit 0, inspected output |
| Behavior is X | Reproduction plus observed output |
| Code matches spec | Diff or path + line range |
| Bug is fixed | Regression fails without, passes with |
| Shipped | All above plus commit or PR link |
Prose and code review are inspection, not verification.
Workflow
- Name the check before editing.
- Run it and show the relevant output; do not paraphrase.
- Inspect the exit code: non-zero makes the claim false. "0 tests", skipped tests, or hidden warnings do not pass.
- Cite the command, output, path, line range, SHA, or test artifact.
- If it fails, fix it or report the work as blocked; do not claim completion.
Red Flags and Rationalizations
"It should work", "I've tested it", "tests pass", "LGTM", "one-line change", "existing tests cover it", "CI will catch it", "I'm in a hurry", and "it's obvious" are not evidence. Run the check again.
Report
Report verification in normal, concise prose; do not emit a mandatory XML/JSON wrapper. Match detail to the claim:
- Result: say whether the requested result is verified, partial, blocked, or unverified.
- Evidence: name the exact command or probe, exit status, relevant output, and path/SHA when applicable.
- Limits: state skipped checks, remaining risks, or unavailable review; do not imply stronger proof than the evidence supports.
A one-line change may need only a diff review and focused command. A code change needs the relevant test/typecheck output. Research needs cited sources and a clear separation of fact, inference, and uncertainty. If no check ran, name the exact intended command and say why; urgency or small scope does not waive verification. Do not claim completion from inspection alone.
04 同仓库更多 Skills
writing-skills
Use when creating new skills, editing existing skills, or verifying skills work before deployment - applies TDD to process documentation by testing with subagents before writing, iterating until bulletproof against rationalization. Includes complete pressure testing methodology.
typescript-coding-standards
Use when writing, refactoring, or reviewing TypeScript code that needs strong domain modeling, typed errors, schema parsing, safe adapters, test seams, or maintainable module boundaries.
test-driven-development
Use when implementing a feature, bugfix, or behavior-changing refactor — require a failing behavior test before production code, then make the smallest change that passes.
source-driven-development
Use when unfamiliar libraries, dependency internals, external APIs, framework behavior, migrations, or current ecosystem guidance require official documentation, version-matched source, read the tests, and cited evidence.
security-and-hardening
Use when auditing security, implementing authentication or authorization, handling secrets, or validating data across external trust boundaries.
prototype
Use when the user wants to prototype, sanity-check a data model or state machine, mock up a UI, explore design options, or says "prototype this", "let me play with it", "try a few designs".