ss-workflow-restrict-git-commands

To prevent destructive git commands from running automatically, you need an authorization policy that requires explicit user confirmation before executing patterns like force push, reset --hard, or clean -f. This Skill enforces per-command consent for these specific high-risk operations.

Destructive git operations to safeguard during agent workflows include push (with force variants), reset --hard, clean -f, branch -D (force deletion), and checkout . / restore . These operations can irreversibly rewrite history or delete local work without explicit confirmation.

Yes, you can block dangerous git patterns at the Claude Code harness level using a PreToolUse hook. This Skill provides an optional deterministic enforcement script that hard-blocks dangerous git operations unless explicitly allowed by the user.

To set up git safety guardrails for repository maintenance or refactoring, activate this Skill at session start by instructing your agent that destructive git operations require explicit authorization, enforcing per-command consent for irreversible changes.

Yes, you need jq installed as a dependency to enforce workflow policies for git authorization. This lightweight command-line JSON processor is required to support the consent validation scripts within the agent harness.

A limitation of per-command consent for destructive git operations is that it requires manual approval for every flagged invocation, which can interrupt automated refactoring or deployment preparation workflows if the agent triggers multiple sequential git commands.