anti-panic-protocol_skill

Overview

This skill enforces a strict, safe failure-handling protocol for tool and command failures under pressure. It limits retries, translates raw errors into clear user-facing messages, and escalates with concrete next steps to avoid collateral damage. The goal is predictable, auditable error responses that protect external systems and user expectations.

How this skill works

On detecting a failure the skill summarizes the failure in one sentence, performs up to two safe, bounded remediation attempts, then stops retries and escalates clearly if unresolved. It verifies outcomes before reporting success and never exposes raw stack traces or internal payloads. The escalation message always states what failed, what was tried, and what the next action is.

When to use it

• A command or external tool fails and immediate retries could cause duplicates or rate-limit violations.
• You need a strict, auditable error-response workflow for automation or agent actions.
• A transient failure occurs but uncontrolled retrying would risk spam or additional failures.
• Authentication, security boundaries, or destructive actions are involved and require escalation.
• You must present a concise status to users or operators without revealing internal diagnostics.

Best practices

• Limit remediation attempts to two unless the user explicitly requests more.
• Always provide one outward message per outcome; do not send duplicate notifications.
• Never include raw stack traces, CLI dumps, or internal tool payloads in user-facing text.
• Respect rate limits and cooldowns; if rate-limited, schedule the next allowed attempt.
• Verify the real-world effect (message sent, file written, event created) before marking success.

Example use cases

• A publish API returns a rate-limit error: perform a single immediate retry, attempt a delayed retry, then escalate with the planned next slot.
• An event creation fails due to date validation: normalize the date and timezone, then ask the user to confirm if unresolved.
• A login flow times out: try token-based auth then a browser flow restart, and request user OAuth confirmation if still blocked.
• A file write to backup storage fails: attempt a safe alternate path, then escalate with the exact error summary and next action.
• An outgoing notification fails during high load: stop retries to avoid duplicates and report the failure with remediation history.