Skill Self Improving Agent
v1.0.0Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⭐ 0· 62·1 current·1 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
Name/description match code and docs: the skill provides reminders, logging templates, hook handlers, and helpers to extract learnings into new skills. All required files (hooks, activator, error-detector, extract helper) are coherent with the stated goal of capturing and promoting learnings.
Instruction Scope
Runtime instructions and scripts stay within the stated scope: creating/maintaining .learnings files, injecting a virtual reminder at bootstrap, detecting command errors via CLAUDE_TOOL_OUTPUT, and scaffolding new skills. The hooks and scripts operate on workspace files and output reminders; they do not read arbitrary system credentials or call external endpoints.
Install Mechanism
No automatic install spec in the registry (instruction-only), and included scripts/handlers are provided in the package. Manual install suggestions reference a GitHub repo (reasonable). There are no downloads from unknown or shortened URLs and no extract-from-remote steps in the registry package itself.
Credentials
requires.env declares none, but scripts and docs expect CLAUDE_TOOL_OUTPUT (used by error-detector.sh). That environment variable is not a secret but is an un-declared runtime input; otherwise no credentials, keys, or unrelated env vars are requested.
Persistence & Privilege
The skill is not force-enabled (always:false) and does not request elevated or persistent global privileges. Hooks/scripts are opt-in and run with the same permissions as the agent process; extract-skill.sh writes files under the current workspace and includes validation to avoid absolute paths or '..' traversal.
Assessment
This skill appears to do what it says: inject lightweight reminders, detect obvious command errors, and scaffold promoted learnings into new skills. Before enabling it: 1) Inspect the included scripts (hooks/openclaw/handler.*, scripts/*.sh) yourself to confirm you’re comfortable with what they write/print; 2) Note that error-detector.sh reads CLAUDE_TOOL_OUTPUT (not a secret) and will output a short reminder when an error pattern is found; 3) Only enable the OpenClaw/Claude hooks if you want those scripts to run with your agent's permissions; consider enabling at project-level rather than globally; 4) Use extract-skill.sh with --dry-run first to see what it would create; 5) Verify the referenced GitHub repo/source if you plan to clone from it. If you need deeper assurance, ask the maintainer for a signed release or run the scripts in a sandboxed workspace first.Like a lobster shell, security has layers — review code before you run it.
latestvk979nxk4v11cgghy2m54c917q184pgf3
License
MIT-0
Free to use, modify, and redistribute. No attribution required.