Panic Prevention Protocol

Prevents panic-driven errors by enforcing a calm, step-by-step recovery process after mistakes or critical feedback to ensure safe, trusted fixes.

This skill is documentation-only and appears to do what it says: teach agents to pause and follow process after critical feedback. The main operational risk is administrative: the files instruct operators to add protocol text to agents' system prompts and install the skill on all agents. Before deploying widely, do the following: 1) Review the SKILL.md and README content yourself to ensure wording and emergency-path handling meets your policies. 2) Test in a staging environment or on a non-production agent to verify behavior against the included test scenarios. 3) Limit initial rollout to agents that actually need this behavior (e.g., manager/devops agents), not every agent by default. 4) Ensure emergency/active-leak paths are well-defined so urgent incidents aren't blocked by wait steps. 5) Monitor agent logs and approvals after rollout to detect unintended overrides. The scanner's 'system-prompt-override' hit is expected given the installation guidance — treat it as a reminder to audit any system-prompt changes rather than a sign of covert exfiltration.

Type: OpenClaw Skill Name: panic-prevention Version: 1.0.5 The 'panic-prevention' skill bundle is a safety-oriented behavioral protocol designed to prevent AI agents from taking impulsive or unverified actions after making mistakes. It consists entirely of markdown documentation and metadata (SKILL.md, README.md, PROTOCOL-FLOW.md) that instructs the agent to follow a structured 9-step recovery process (S.B.A.P.P.W.E.V.D.) emphasizing human approval and process adherence. There is no executable code, no network activity, no data exfiltration, and no evidence of malicious prompt injection; rather, the instructions serve as a defensive guardrail to increase human oversight during critical failures.