Self Improvement
v0.1.0Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⭐ 0· 97·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (capture learnings, log errors, promote learnings) match the files and behavior. The hook, activator, error detector, and extract helper are coherent with the stated purpose. No unrelated credentials, binaries, or external services are requested.
Instruction Scope
SKILL.md instructs installing files into ~/.openclaw and enabling an OpenClaw hook that injects a reminder into every session's bootstrap context — this is explicitly the feature (workspace-based prompt injection for reminders). The PostToolUse hook runs a shell script that reads CLAUDE_TOOL_OUTPUT to detect errors; that is expected for an error-detection hook but means the script will see tool outputs. The skill's instructions do not ask to read unrelated system files or to transmit data to external endpoints.
Install Mechanism
There is no automatic installer; SKILL.md recommends git clone or a ClawdHub install. No remote binary downloads or archive extraction are used. The provided scripts/hook files are local and standard; the extract script writes only into a relative ./skills directory and includes checks to prevent absolute paths or '..' traversal.
Credentials
The skill declares no required environment variables or credentials. The error-detector script reads CLAUDE_TOOL_OUTPUT (documented in references) which is appropriate for PostToolUse hooks; no other env or secret access is present.
Persistence & Privilege
always:false (no forced global inclusion). Enabling the OpenClaw hook is explicit and will inject reminder text into session bootstrap files when turned on — this affects every session the hook is enabled for and can influence model behavior (intended functionality). The only file-writing behavior is the extract-skill.sh helper (creates a skill scaffold in a relative directory) and instructions to create ~/.openclaw/workspace/.learnings; these are reasonable for the purpose but should be authorized by the user.
Assessment
This skill appears to do what it says: remind agents to log learnings and optionally detect failed commands and scaffold new skills. Before installing or enabling hooks: 1) Review the hook and script files (hooks/openclaw/handler.{js,ts}, scripts/*.sh) to ensure you trust their content and the repo source; 2) Understand that enabling the hook injects a reminder into every session bootstrap (this intentionally modifies session context and can influence assistant behavior); 3) If you enable the PostToolUse hook, be aware the error-detector reads CLAUDE_TOOL_OUTPUT (tool outputs may include sensitive data), so limit that hook to project-level settings or add matcher filters if you don't want global activation; 4) The extract-skill.sh will create files under a relative ./skills path — run with --dry-run first and ensure file permissions are appropriate; 5) Prefer installing in a controlled user environment (project-level settings) rather than globally, and verify the upstream GitHub repository if you want an auditable source.Like a lobster shell, security has layers — review code before you run it.
latestvk978dxjbwf7ysevj977p9cem1583e2wn
License
MIT-0
Free to use, modify, and redistribute. No attribution required.