What problem does it solve? Autonomous coding agents can silently bypass enforcement hooks, push to shared branches, overwrite immutable result artifacts, or sweep unrelated files into commits. This Skill defines the blocking rules and review checks that keep agent behavior inside the repository's safety boundaries. ## Core Features & Use Cases - Destructive-operation blocking: Blocks or requires explicit confirmation for git push to shared branches, git reset --hard, force-pushes, git commit --no-verify, and deletion of result artifacts. - Environment review checklist: Verifies pre-commit contract gates, write-time hooks (notation guard, results no-overwrite, dispatch-readiness guard), staged-file scope, and worktree setup before dispatch. - Escalation protocol: Surfaces conflicts between guardrails and user instructions instead of silently obeying either side. - Use Case: Before dispatching an autonomous agent on a research repository, run the required checks to confirm hooks are firing, the worktree has .env copied in, and the dispatch prompt states hard stops and the results/ provenance rule. ## Quick Start Review my current agent environment against the safety guardrails and confirm all enforcement hooks are active before I dispatch an autonomous task.
