Self Improving Agent
v3.0.15Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⭐ 0· 124·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
Capability signals
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (capture learnings, errors, promote learnings) align with the included files: README-like SKILL.md, templates, log formats, an OpenClaw hook handler, and small helper scripts. Required env/config/credentials are none, which is appropriate for a logging/templating skill.
Instruction Scope
Instructions ask you to create .learnings/ files, copy hook files into ~/.openclaw/hooks, enable the hook, and optionally wire scripts into agent hook events. That scope matches the stated purpose. Note: the error detector script reads CLAUDE_TOOL_OUTPUT (a platform-provided variable) to detect failures — expected for its purpose. None of the instructions request unrelated system files or credentials, but enabling hooks gives the skill the ability to inject context into sessions (opt-in).
Install Mechanism
There is no packaged install spec; the docs recommend git clone from a GitHub repo or using ClawdHub. All bundled scripts/handlers are local text files. No remote downloads, URL shorteners, or extract/install of arbitrary archives are used by the skill itself.
Credentials
The skill declares no required environment variables or credentials. The scripts expect platform-provided variables (e.g., CLAUDE_TOOL_OUTPUT) and normal filesystem access for creating log files; this is proportional to a logging/hook skill.
Persistence & Privilege
always:false and no automatic global takeover. However, the user-facing instructions encourage installing hooks at user-level (e.g., ~/.claude/settings.json or enabling OpenClaw hooks), which would make the activator/error-detector run for every session you opt into. That is an opt-in persistent behavior and should be enabled deliberately. The hook handler injects a virtual bootstrap file (reminder) — expected behavior for this skill.
Assessment
This skill appears coherent and does what it says: log learnings and optionally inject reminders via hooks. Before installing: review the hook and scripts (hooks/openclaw/handler.{js,ts}, scripts/*.sh) to confirm you trust the code and the originating repo; prefer installing per-project rather than user/global scope if you want limited exposure. If you enable the hooks globally or add the activator to a user-level settings file, the scripts will run for every session (with the same permissions as your agent) — ensure you’re comfortable with that. The extract-skill.sh script creates files under your workspace (it contains safeguards against writing outside the current directory), but inspect it before running. If uncertain, try the skill in a sandboxed project or repository first.Like a lobster shell, security has layers — review code before you run it.
latestvk97c81r244w3d0syrc0fas399x84ew7m
License
MIT-0
Free to use, modify, and redistribute. No attribution required.