27
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill saadshahd/moo.md --skill gate- SKILL.md4.0 KB
Overview
This skill enforces a verification gate before claiming work is done, fixed, or complete. It runs a checklist tailored to workflow type (build, debug, refactor) and requires concrete evidence rather than assumptions. Use it to reduce regressions, unclear handoffs, and overconfident completion claims.
How this skill works
Before work begins, the Pre-Work Gate ensures a locked verification method, at least two explicit "must NOT" criteria, and a fit score that shapes the workflow. After work, the Post-Work Gate runs workflow-specific checks (Build, Debug, Refactor) and requires execution output, observation, or measurement as proof. The skill blocks completion claims if required evidence or checks are missing and provides a clear incomplete-work template when blocked.
When to use it
- Right before declaring work "done", "fixed", or "complete"
- When preparing a PR, build, or deploy that affects production
- After debugging to confirm root cause and prevention
- When refactoring to ensure behavior unchanged
- During planning to set verification approach and avoid assumptions
Best practices
- Lock a verification method that yields execution output, observation, or measurement—avoid "assumption"
- Define at least two explicit "must NOT" constraints to limit scope creep
- Choose and document a fit score to determine checkpoints and reviewer involvement
- For high-stakes changes require both execution output and observation
- Show actual test or command output when claiming tests passed; screenshots or metrics are valid evidence
- If verification exceeds ~15 minutes, split the task—long verification is a sign the work is too big
Example use cases
- Feature build: run feature, show output, exercise edge cases, verify dependencies and rollback path
- Bug fix: capture root cause evidence, run fix that resolves the symptom, add a prevention test or guard
- Refactor: run unchanged tests, produce output diffs to prove identical behavior, remove orphaned code
- Pre-work: decide verification type and evidence before starting a change that touches multiple services
- Triage: when unsure, block and publish what's done, what's blocking, and next steps with evidence
FAQ
Execution output, observation (screenshots/debugger), or measurement (metrics/benchmarks). Code review alone is weak; assumptions block shipping.
When is a quick change allowed without full gate?
Type 2A exception: trivial changes with <1 minute rollback may be claimed if you document the undo command and why the change is trivial.