Detect AI agent task deviations and execute a structured course correction process.
npx skills add https://github.com/gweslab/cerf --skill bad-gweslab
Please help me install this Agent Skill. Skill: bad Source: https://github.com/gweslab/cerf/tree/main/.claude/skills/bad Command: npx skills add https://github.com/gweslab/cerf --skill bad-gweslab
The Skill helps you identify and correct when your AI agent deviates from the task or rules, ensuring the agent follows the intended plan accurately.
/bad to correct the trajectory.If your AI agent has drifted from the task or violated the rules, use the /bad command to reset its trajectory.
High-intent search queries and answers about installing and using this skill.
To course correct an AI agent when it drifts from the task or rules, you invoke the `/bad` command to stop the wrong trajectory, articulate specific violations, propose a new plan, and execute the recovery process.
The recovery process for an incorrect work trajectory involves stopping the agent immediately, articulating specific rule violations with concrete examples, proposing a new plan, reverting bad work, and executing the corrected plan.
To stop your AI agent from writing incorrect code and reset its task adherence, you invoke the `/bad` command to force the agent to articulate violations, propose a strict new plan, and execute a structured course correction.
You should trigger a course correction for your AI agent when you identify that it has deviated from the intended task or violated documented rules, ensuring it stops working on the wrong trajectory immediately.
Yes, the AI agent course correction process requires the agent to follow specific rules and tasks as documented, ensuring the agent can articulate specific violations and propose a new plan that adheres strictly to the original task.
dnhessAudit agent work for intent, consistency, constraints, and trajectory coherence.
erfanzarDispatch implementer subagents per task with two-stage spec and quality reviews.
0xb11aMonitor open crypto positions and trigger automatic exits on danger signals.
FatByEnable AI agents to anticipate needs and self-correct using WAL Protocol and Working Buffer.