Self-Improving Experience Memory
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md to install @user_aba7511a/self-improving-skills into your AI assistant.
About this skill
Problem: Agent Experience Is Not Durable
AI agents often encounter command failures, unexpected API behavior, user corrections, and better workflows, but that knowledge usually stays inside one conversation. Without external records, the next session may repeat the same mistakes, re-explain the same constraints, or miss context that was already discovered. self-improving-Skills addresses this fragmentation by turning errors, corrections, feature requests, and knowledge gaps into structured Markdown entries that can be searched, linked, and promoted over time.
How It Works: From Logs to Memory
The skill routes different signals into separate files: command or external tool failures go to .learnings/ERRORS.md; user corrections, knowledge gaps, and best practices go to .learnings/LEARNINGS.md; feature requests go to .learnings/FEATURE_REQUESTS.md. Entries use IDs in the form TYPE-YYYYMMDD-XXX, which makes retrieval and cross-referencing easier. When a similar issue appears again, it searches existing entries, adds a See Also link, increments Recurrence-Count, and promotes the lesson into WorkBuddy memory files such as ~/.workbuddy/MEMORY.md or .workbuddy/memory/MEMORY.md when thresholds are met. It also prompts a short post-task review: were there mistakes, corrections, or more reliable workflows worth recording?
Boundaries
This skill fits long-running agent workflows, especially when the agent repeatedly handles similar tasks, external tools, or user corrections. It is not a replacement for tests, CI, monitoring, or incident response. It should not record secrets, tokens, private keys, or environment variable values. If the issue is a one-off configuration mistake, a normal debugging note may be enough; if the workflow is reusable and executable, it should be turned into a proper Skill.
Use Cases
- Debugging recurring external API timeouts and capturing errors, responses, and troubleshooting steps as searchable entries.
- Recording repeated user corrections about output format so the rules can be reviewed and reused in later tasks.
- Distilling recurring project conventions into WorkBuddy workspace memory files while maintaining an agent.
- Reviewing failed tool calls after a task to decide whether to mark entries pending, promote memory, or create a Skill.
Best For
- AI agent engineers who want to capture runtime errors and user corrections as reviewable Markdown records.
- WorkBuddy assistant builders who need to promote cross-session lessons into user or workspace memory.
- Technical leads reviewing agent quality who track recurring issues, entry status, and Skill candidates.
- Developers debugging external tool chains who record command failures, API behavior mismatches, and sanitized clues.
Related Skills
Automatically searches job postings based on the user profile, AI-scores fit, saves desktop reports, and sends email updates with scheduled tracking.
Activates bionic reasoning for causal judgment, numerical prediction, and hypothesis validation, using hypothesis-driven checks, Bayesian updates, falsifiability tests, bias defense, and physical constraints.
Triggered by /plan, it asks the agent to output a plan, risks, impact scope, and validation approach before execution.
Track and clean agent session files, packages, and Skills via trash-first safety.