← ClaudeAtlas

hardenlisted

Use when asked to make an AI or LLM workflow resilient, add a guard, policy check, or safety boundary around an irreversible action, prevent an agent from ever doing something, or fix a finding from a resilience review.
faultkit/skills · ★ 0 · AI & Automation · score 73
Install: claude install-skill faultkit/skills
Read `${CLAUDE_PLUGIN_ROOT}/faultkit/SKILL.md` and follow its **Harden** mode for: $ARGUMENTS Start by counting the invariants. When more than one is unguarded, ask whether to harden them one at a time, all in a row, or on the user's instruction, and wait for the answer. End by asking whether to run faultkit to prove it, then whether to open a pull request, waiting for each answer. If `$ARGUMENTS` contains `--auto`, do not ask: harden every unguarded invariant, continue into Prove or Prove all, and print the pull request commands instead of opening one. The safety gate still runs first.