Self Improving Agent Jarvis
v3.0.11Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⭐ 0· 31·0 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
The name/description (capture learnings/errors/promote to workspace) aligns with the included artifacts: logging templates, hook handlers that inject reminders, activator/error-detector scripts, and a skill-extraction helper. Required env vars, binaries, and config paths are not requested, which matches the stated lightweight purpose.
Instruction Scope
SKILL.md and hook scripts instruct the agent to create and update local workspace files (e.g., ~/.openclaw/workspace/.learnings and AGENTS.md/SOUL.md). The error-detector script reads the CLAUDE_TOOL_OUTPUT env var (a platform-provided variable) to detect failures; this env var is not listed in requires.env (it's platform-provided), so be aware the hooks examine tool output. All instructions are focused on logging/promoting learnings and enabling reminders—no broad or unrelated data collection or exfiltration is described.
Install Mechanism
No automated install spec is present (instruction-only). Manual install suggestions use a GitHub repo and a ClawdHub command; scripts live in the repo. There are no remote downloads from arbitrary URLs or extracted archives. Installing requires cloning/copying files into the user's OpenClaw directories (user action required).
Credentials
The skill declares no required credentials or secrets. The only environment access observed in scripts is reading CLAUDE_TOOL_OUTPUT (used to detect command errors) and normal filesystem paths under the user's home or current workspace; both are proportionate to a logging/reminder skill.
Persistence & Privilege
The skill provides hooks that the user can copy into ~/.openclaw/hooks and enable; when enabled these will run on agent lifecycle events and inject reminders into sessions. 'always' is false and the skill is opt-in, but enabling hooks gives the skill persistent runtime presence across sessions (normal for hook-based workflows). Review/enable intentionally.
Assessment
This skill appears to do what it says: add lightweight reminders, detect command errors, and provide templates to log learnings. Before installing or enabling hooks: 1) Review the scripts (activator.sh, error-detector.sh, extract-skill.sh) yourself — they will run with the same permissions as your agent; 2) If you enable the hook, it will be invoked on agent events and will inject virtual files / output into sessions — enable only where you want persistent reminders; 3) The error detector reads CLAUDE_TOOL_OUTPUT (a platform-provided env var) to spot failures — ensure your agent/platform uses that var and that you’re comfortable with hooks reading tool output; 4) The extract-skill helper writes files into the current workspace (it tries to prevent writes outside the workspace) — run with care and check the output paths. If you don't use OpenClaw/Claude hooks, you can still use the templates manually without enabling hooks.Like a lobster shell, security has layers — review code before you run it.
latestvk97aeg2mtzc6w270he1ynnet5184320j
License
MIT-0
Free to use, modify, and redistribute. No attribution required.