What problem does it solve? Long AI-assisted coding sessions suffer from "Context Rot" — quality degrades as the agent processes more information, leading to repeated failed debugging, circular reasoning, and guesswork. This Skill gives the agent self-monitoring rules to detect these failure patterns and preserve state before quality collapses. ## Core Features & Use Cases - 3-Strike Rule: After 3 failed debugging attempts on the same issue, the agent stops, documents what was tried in .gsd/STATE.md, and recommends a fresh session instead of continuing to flail. - Circular Detection: Recognizes when the same approach is being retried, lists prior attempts, and proposes a fundamentally different approach or a /pause. - Uncertainty Logging: Records uncertain decisions in .gsd/DECISIONS.md with alternatives considered, prompting user guidance rather than guessing. - Use Case: During a long debugging session, the agent fails to fix a test three times. Instead of trying a fourth variation, it writes a structured state dump with attempts, errors, and hypotheses, then hands off cleanly via /pause and /resume. ## Quick Start Ask the agent to monitor its own context health and write a state dump to .gsd/STATE.md if it repeats failed debugging attempts.
