my-test
v1.0.1Captures learnings, errors, and corrections to enable continuous improvement. Use when: (1) A command or operation fails unexpectedly, (2) User corrects Clau...
⭐ 0· 261·0 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 skill's stated purpose is to capture learnings/errors and promote them into workspace files or extracted skills. The included files (SKILL.md, hook handlers, activator/error-detector scripts, extract helper) directly implement that purpose. The registry name/slug ('my-test') differs from the skill frontmatter/name ('self-improvement' / 'self-improving-agent'), which is a cosmetic inconsistency but not a functional mismatch.
Instruction Scope
SKILL.md and the hook/docs explicitly instruct creating ~/.openclaw/workspace/.learnings, copying hooks into ~/.openclaw/hooks, and optionally cloning from GitHub. The activator only emits a small reminder. The error-detector reads CLAUDE_TOOL_OUTPUT to detect command failures and emits a reminder on error. The extract-skill.sh script creates files under a ./skills directory (with path validation). These behaviors are consistent with the stated goal, but the instructions do require writing files (hooks, workspace files, extracted skills) and running scripts that will run with the user's agent permissions — review those scripts before enabling them.
Install Mechanism
There is no automated install spec in the registry (instruction-only), and the README suggests a manual git clone from a GitHub repo. All code files are present in the package (hooks and shell scripts). There are no downloads from unknown hosts, no URL shorteners, and the provided scripts validate paths before writing. This install mechanism is proportionate to the skill's purpose but requires you to run commands that write to your filesystem if you follow the instructions.
Credentials
The skill declares no required environment variables or credentials. At runtime the error-detector reads CLAUDE_TOOL_OUTPUT (a runtime-provided environment variable for the tool output) and hook handlers inspect event context; these are appropriate for a PostToolUse / agent:bootstrap integration. There are no requests for unrelated secrets or system credentials.
Persistence & Privilege
always is false and the skill is user-invocable. The hooks will only run if you explicitly copy/enable them (instructions show cp and openclaw hooks enable). Once enabled, the hook injects reminder content into every session bootstrap (i.e., it influences session context). This is expected for a workspace-injection style skill, but be aware enabling hooks gives the code the ability to change session bootstrap state for all sessions — enable per-project or review code to limit scope if desired.
Assessment
What to check before installing/enabling:
- Confirm the repo/source before running git clone or copying files into ~/.openclaw or ~/.claude; the SKILL.md references a GitHub project — review that upstream repository.
- Inspect the provided scripts (activator.sh, error-detector.sh, extract-skill.sh) yourself. activator.sh and error-detector.sh only emit reminders (safe), but extract-skill.sh creates files under a skills directory (it validates paths but will write files when run).
- Enabling hooks will inject content into every session bootstrap; prefer per-project enablement instead of global user-level enablement if you want to limit blast radius.
- There are no requested credentials, but the hooks/scripts run with the same permissions as your agent process: ensure they are acceptable to run under that account and check file permissions before enabling.
- Consider using the script's --dry-run option (extract-skill.sh supports --dry-run) and test in an isolated workspace first.
- The mismatch between the registry entry name (my-test) and the SKILL.md name (self-improvement/self-improving-agent) is a minor red flag—verify you are installing the intended skill.Like a lobster shell, security has layers — review code before you run it.
latestvk97drjkdzd5j8gjjbr1naj1nsx82nxe9
License
MIT-0
Free to use, modify, and redistribute. No attribution required.