Log learnings, errors, and corrections to improve AI agent behavior.
npx skills add https://github.com/FatBy/DD-OS --skill self-improvement-fatby
Please help me install this Agent Skill. Skill: self-improvement Source: https://github.com/FatBy/DD-OS/tree/main/skills/self-improving-agent Command: npx skills add https://github.com/FatBy/DD-OS --skill self-improvement-fatby
This Skill enables AI agents to learn from their mistakes, user corrections, and knowledge gaps, leading to continuous improvement and more robust performance over time.
Log a user correction to the learnings file by stating 'User corrected me on the date format, it should be YYYY-MM-DD'.
High-intent search queries and answers about installing and using this skill.
You enable continuous improvement for AI agents by logging command failures, user corrections, and knowledge gaps into a structured file. This allows the agent to capture learnings and avoid repeating the same mistakes in future operations.
You log user corrections by explicitly stating the correction, such as noting a specific date format requirement. The skill captures this feedback to refine agent behavior and record the correct approach for future tasks.
Use agent self-correction when a command fails due to incorrect parameters or unexpected outcomes. The mechanism automatically logs the error and the correct usage, allowing the agent to learn and adapt its execution strategy.
Yes, you can identify outdated knowledge gaps by flagging instances where the agent's knowledge is incomplete. The skill records these gaps to facilitate adaptive learning and enhance the overall knowledge base.
The best way to promote successful strategies is using the skill's promotion mechanisms to elevate learnings to higher-level documentation. This ensures broader agent use and continuous improvement across various scenarios.
acarmonagCoordinate agent workflows with MCP-integrated hooks for validation and session persistence.
akiyoshiAggregate and search session learnings stored in learnings.jsonl.
HafizFasihMonitor system health across watchers and MCP servers with automatic recovery.
wineastExtracts lessons from conversations and stores them as organized rules in CLAUDE.md and .claude/rules/ files.
