Temp Skill Download
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· 104·1 current·1 all-time
byBoBoinChina@boboincn
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (self-improvement / logging learnings) matches the included assets: templates, log formats, hook handlers, activator/error-detector scripts, and an extraction helper. There are no requests for unrelated credentials, binaries, or system config.
Instruction Scope
SKILL.md and reference docs instruct creating .learnings/ files, copying the hook into ~/.openclaw/hooks, and optionally cloning a GitHub repo. The scripts read CLAUDE_TOOL_OUTPUT (documented) to detect errors. All referenced file paths and actions (writing to workspace, adding bootstrap virtual file, running local scripts) are consistent with the stated logging/promotion behavior. Minor note: the docs recommend cloning from GitHub and enabling hooks — these are normal for installation but are actions that write to and modify the user's workspace if followed.
Install Mechanism
There is no automated install spec in the package (instruction-only), and included scripts/hooks are present in the bundle. The SKILL.md suggests manual git clone or use of a 'clawdhub' installer; cloning from GitHub is standard but is an out-of-band action the user must opt into. No external, opaque download URLs or archive extraction are present in the package itself.
Credentials
The skill does not declare required env vars or credentials. The scripts use optional env vars (e.g., SKILLS_DIR override) and read CLAUDE_TOOL_OUTPUT to detect errors — both are proportional to the purpose. There are no requests for secrets or unrelated service tokens.
Persistence & Privilege
always:false and model invocation enabled (defaults) — normal. The OpenClaw hook, if installed/enabled by the user, will inject a virtual reminder at agent bootstrap and the activator/error-detector will run when configured. The extract-skill.sh can create files under a skills directory when executed. These behaviors require explicit user configuration/enabling; review before enabling hooks globally.
Assessment
This skill appears to do what it says: remind the agent to log learnings and provide helpers to promote those learnings into workspace files or extracted skills. Before installing or enabling hooks: 1) Review the scripts (activator.sh, error-detector.sh, extract-skill.sh) to ensure you understand where they write files (defaults: ~/.openclaw/workspace/.learnings or ./skills) and that the SKILLS_DIR override is acceptable. 2) Hooks run with the same permissions as your agent — only enable them if you trust the code and repository. 3) The error detector reads CLAUDE_TOOL_OUTPUT to look for error strings; it only outputs reminders (no network calls), but it may trigger frequently if enabled globally. 4) If you run extract-skill.sh, it will create files/directories—use --dry-run first or set an explicit --output-dir. 5) If you want tighter control, enable hooks at project-level (not user/global) and/or inspect the repository you clone. Overall, no credentials or external endpoints are requested and no obfuscated or suspicious code was found.Like a lobster shell, security has layers — review code before you run it.
latestvk979grh0z2v9tc3vcspd1d761x83bj1m
License
MIT-0
Free to use, modify, and redistribute. No attribution required.