Skill

Smart skill installation advisor for ClawHub. Searches for skills matching your needs, evaluates candidates on security (via skill-shield), code quality, and...

This skill is an advisor that will use your clawhub CLI to install and inspect other skills and will run a local skill-shield scanner. That is consistent with its purpose but: (1) clarify the metadata mismatch — it DOES need clawhub and a scanner even though metadata claims 'zero external dependencies'; (2) only run it in a disposable or isolated workspace (use --workdir pointing at a temp directory or container) because 'clawhub install' can run package install hooks from third-party skills; (3) verify the scanner path (SKILL_SHIELD_SCANNER) and review the generated report before allowing any recommended installation; (4) if you need higher assurance, run the tool with --top 1 and inspect the candidate skill contents manually before letting it install or recommend anything; (5) if the author can explain/correct the contradictory metadata (required binaries/env vs declared none), that will raise confidence.

Type: OpenClaw Skill Name: skill-store Version: 2.3.1 The 'skill-store' skill, designed for security evaluation, contains critical vulnerabilities. The `scripts/evaluate.py` script allows arbitrary code execution via the `--scanner` argument and arbitrary file write/deletion via the `--workdir` argument, due to insufficient input validation when handling user-provided paths. Additionally, the `webapp/index.html` is vulnerable to Cross-Site Scripting (XSS) as it directly interpolates skill names and descriptions from `skills.json` into the DOM without proper sanitization. While these flaws could enable severe attacks, there is no clear evidence of intentional malicious behavior by the 'skill-store' itself; rather, they are vulnerabilities in its implementation.