| name | verifier_hooks |
| description | Arm a one-shot hook-driven verifier for the next risky edit. Use when Claude should automatically run a lightweight post-edit agent check after the next Edit or Write and stop the turn if the change likely introduced a concrete regression. |
| allowed-tools | ["Bash","Read","Grep","Glob"] |
| disable-model-invocation | true |
| metadata | {"version":"2.0","effort":"low","auto-invocable":false,"category":"workflow","compatible-claude-code":{"when_to_use":"When arming a one-shot verifier for risky edits","allowed-tools":["Bash","Read","Grep","Glob"]}} |
| hooks | {"PostToolUse":[{"matcher":"Edit|Write","hooks":[{"type":"agent","prompt":"You are a lightweight post-edit verifier. The hook input JSON is:\n$ARGUMENTS\n\nInspect the touched file and nearby code using read-only tools only.\nReturn ok=true unless there is a concrete reason to stop the workflow immediately.\n\nFail with a short reason only when one of these is true:\n1. the edit likely broke syntax, imports, or obvious runtime behavior\n2. the edit changed auth, security, migrations, deployment, or persistence logic without a targeted verification path\n3. the edit removed or contradicted an existing nearby invariant or targeted test\n4. the edit changed hooks, skills, agents, or frontmatter in a way that is internally inconsistent\n\nIgnore style-only issues and speculative concerns. Prefer passing if unsure.\n","timeout":45,"statusMessage":"Running edit verifier","once":true}]}]} |
Verifier Hooks Skill
Usage: /verifier_hooks
This skill arms a one-shot PostToolUse agent hook for the next Edit or Write in the current session.
After Arming
- Tell the user the next risky edit is now protected by a lightweight verifier.
- Recommend it for auth changes, deploy logic, migrations, broad refactors, or hook/skill/agent edits.
- If the user wants the verifier on every risky edit, move the same pattern into
.claude/settings.local.json instead of repeatedly invoking this skill.
Rules
- Do not try to remove hooks manually; this one self-removes because
once: true.
- Do not run extra verification immediately inside this skill invocation; the hook is the verification mechanism.
