Manages task execution states and retry logic through structured attempt tracking and automated gating.
The Attempt Ledger skill provides a robust framework for managing complex task execution within Claude Code epics. It tracks every execution attempt, providing the AI with the necessary failure context to perform intelligent retries while implementing strict gating rules to prevent infinite loops and wasted tokens. By utilizing structured receipts for both successes and failures, the skill ensures progress persistence across sessions and facilitates 'compound learning,' where insights from individual tasks are automatically aggregated and promoted to permanent project documentation in AGENTS.md.
Key Features
0118 GitHub stars
02Compound learning extraction that promotes reusable insights to AGENTS.md
03Automated gating rules to prevent infinite retry loops and resource waste
04Persistent task status management (pending, in_progress, completed, gated)
05Seamless progress resumption and state re-anchoring for AI agents
06Structured attempt tracking with failure context and success receipts
Use Cases
01Building a project-specific knowledge base by capturing and aggregating task-level learnings
02Implementing automated quality gates and retry limits for autonomous coding tasks
03Managing multi-step development epics that require reliable task sequence execution
