← Back to Skills Marketplace
Self Optimization
by
Alethean-kaw
· GitHub ↗
· v1.0.2
· MIT-0
144
Downloads
0
Stars
1
Active Installs
3
Versions
Install in OpenClaw
/install self-optimization
Description
Turn mistakes, corrections, dead ends, and repeated fixes into durable improvements. Use when work reveals a non-obvious lesson, a recurring failure, a missi...
Usage Guidance
This skill is internally coherent and appears safe in that it only writes local markdown entries, emits reminders, and scaffolds templates. Before installing: (1) review the scripts (scripts/*.sh) to confirm you are comfortable with local writes; they create files under the workspace and avoid writing outside the current directory; (2) enabling hooks is opt-in — only enable OpenClaw/Codex/Claude hooks if you want the reminders; (3) the error detector reads CLAUDE_TOOL_OUTPUT (tool output provided by the agent runtime) to detect failures — that data stays local unless you explicitly add forwarding; (4) ensure scripts are executable and the workspace .learnings/ location has appropriate permissions; (5) if you prefer not to have automatic reminders, install without enabling the hooks or narrow the hook matcher. Overall, nothing here indicates credential exfiltration or remote code fetching, but always review and only enable hooks you trust.
Capability Analysis
Type: OpenClaw Skill
Name: self-optimization
Version: 1.0.2
The 'self-optimization' skill bundle is designed to help OpenClaw agents learn from errors and user feedback by maintaining a structured log in a `.learnings/` directory. The bundle includes shell scripts (`extract-skill.sh`, `error-detector.sh`) for automation and OpenClaw hooks (`handler.js`) to inject reminders into the agent's context. The code includes appropriate safeguards, such as path traversal checks in the skill extraction script, and the instructions in `SKILL.md` are strictly aligned with the stated purpose of improving agent performance through iterative feedback.
Capability Assessment
Purpose & Capability
Name/description match the provided files. The package includes templates, lightweight hooks, and helper scripts to record and promote learnings; these are appropriate for a "self-optimization" skill.
Instruction Scope
SKILL.md instructs creating and writing to a workspace .learnings/ inbox and promoting entries. The included hooks and scripts are limited to emitting reminders, scanning local tool output for common error patterns, and scaffolding new skill templates. One runtime script reads the CLAUDE_TOOL_OUTPUT environment variable (used to surface tool output) — this is consistent with an error-detector hook but is not declared in requires.env, so it's worth noting.
Install Mechanism
No external downloads or package installs. This is effectively instruction-plus-local-scripts; everything is contained in the repo and scripts are executed locally. No URLs, extract operations, or third-party installers are used.
Credentials
The skill declares no required environment variables or credentials, which is appropriate. The only environment usage observed is reading CLAUDE_TOOL_OUTPUT in scripts/error-detector.sh (expected for a PostToolUse hook). No other secrets, tokens, or unrelated credentials are requested.
Persistence & Privilege
always:false and no attempts to modify other skills or global agent config were found. The hook injects a virtual bootstrap file and scripts create files under the current workspace or a relative ./skills path — this is proportionate for the stated purpose.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install self-optimization - After installation, invoke the skill by name or use
/self-optimization - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.2
self-optimization v1.0.2
- Added standard log files: `.learnings/ERRORS.md`, `.learnings/FEATURE_REQUESTS.md`, and `.learnings/LEARNINGS.md` to structure captured learnings, errors, and feature requests.
- No changes to logic or process; updated workspace to include these files for consistent tracking and review.
v1.0.1
self-optimization 1.0.1
- Removed internal `.learnings/*` log files from the skill package.
- Added a `README.md` for documentation.
- No changes to overall guidance or core workflow.
v1.0.0
Self-Optimization v1.0.0
- Introduces a structured system for capturing and promoting learnings from mistakes, corrections, and recurring issues to prevent repeated failures.
- Defines `.learnings/` directory with dedicated logs for learnings, errors, and feature requests, each with detailed templates.
- Details rules for detecting and prioritizing meaningful learning signals, linking recurrences, and guidelines for when to promote patterns into durable project guidance or reusable skills.
- Includes clear promotion pathways into team documents (like `AGENTS.md`, `SOUL.md`, `TOOLS.md`) and extraction workflow for new skills.
- Provides reference workflows, review checkpoints, and setup instructions for integration with OpenClaw and agent hooks.
Metadata
Frequently Asked Questions
What is Self Optimization?
Turn mistakes, corrections, dead ends, and repeated fixes into durable improvements. Use when work reveals a non-obvious lesson, a recurring failure, a missi... It is an AI Agent Skill for Claude Code / OpenClaw, with 144 downloads so far.
How do I install Self Optimization?
Run "/install self-optimization" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Self Optimization free?
Yes, Self Optimization is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Self Optimization support?
Self Optimization is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Self Optimization?
It is built and maintained by Alethean-kaw (@alethean-kaw); the current version is v1.0.2.
More Skills