From meta
Turn a real agent failure into a permanent guard in the right harness layer (AGENTS.md line, rule, hook, subagent check, or skill). Use when the agent made a mistake you don't want to see again — "the ratchet."
How this skill is triggered — by the user, by Claude, or both
Slash command
/meta:hardening-the-harnessThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The agent did something wrong — shipped broken/“finished” code, ran a destructive command, ignored a convention, derailed on a long task — and you want it to never happen again. Don't just retry; ratchet it.
The agent did something wrong — shipped broken/“finished” code, ran a destructive command, ignored a convention, derailed on a long task — and you want it to never happen again. Don't just retry; ratchet it.
AGENTS.md, or a scoped .mdc rule (use rule-author).harness-hooks runbook). Promote a rule to a hook if the agent keeps violating it.planning-a-change).reviewer subagent check.reviewer / security-reviewer agent.skill-author).AGENTS.md/rules stay short or they get ignored).rule-author, skill-author, learning-from-chatsharness-setup, harness-hooksagent-harness-engineeringnpx claudepluginhub naffis/loadout --plugin metaProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.