ClawHub

self-improving-agent

v3.0.16

Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...

⭐ 3.2k· 395k·5.9k current·6.2k all-time
by@pskoett
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
The name/description (capture learnings, errors, feature requests) matches the included files and behavior: README, SKILL.md, .learnings templates, small helper scripts (activator, error detector, extractor) and OpenClaw hook handlers. No unrelated credentials, binaries, or surprising capabilities are requested.
ℹ
Instruction Scope
Runtime instructions are focused on creating/maintaining .learnings/ files and optionally installing a hook to inject reminders at bootstrap. The error-detector hook reads CLAUDE_TOOL_OUTPUT to detect failures; although the SKILL.md explicitly warns not to log secrets, reading tool output means sensitive data could be observed or recorded if the agent or user chooses to include full transcripts.
✓
Install Mechanism
No automated install spec or remote archive fetch is used. Installation is via Clawdhub or manual git clone / copy into ~/.openclaw, and copying hooks into ~/.openclaw/hooks. These are standard and transparent operations; there are no downloads from shorteners/personal servers or opaque extract steps.
ℹ
Credentials
The skill declares no required env vars or credentials. In practice the error-detector script expects CLAUDE_TOOL_OUTPUT (a hook-provided context variable) and the hooks receive event.context. This is proportionate to an error-detection hook, but it means command/tool output (which can contain secrets) is accessible to the skill when hooks are enabled.
✓
Persistence & Privilege
The skill is not always: true and does not request forced global inclusion. Hook installation and enabling is opt-in (user must copy/enable hooks). The hook injects a virtual reminder file and avoids overwriting existing files; the skill does not modify other skills' configs or demand system-wide privileges.
Assessment
This skill appears coherent and implements what it claims: logging learnings and providing optional hooks for reminders and basic error-detection. Before enabling it: (1) Review the provided scripts (activator.sh, error-detector.sh, extract-skill.sh) yourself to confirm they match your expectations; (2) Be cautious enabling the PostToolUse / error-detector hook globally — it reads tool output (CLAUDE_TOOL_OUTPUT) which can contain secrets or sensitive command output; prefer enabling hooks only in trusted/workspace-scoped environments; (3) Follow the SKILL.md guidance not to record secrets — if you ask the agent to include full transcripts or raw command output, sensitive data may be persisted; (4) If you plan to install via git clone, verify the repository URL and contents; (5) Test in a disposable project or safe workspace before enabling system- or user-level hooks.

Like a lobster shell, security has layers — review code before you run it.

latestvk97d424jyd3pag4d6q7rzm0rkn84yxze

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments