text-cleaner

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill's required "Auto-Update Check" (SKILL.md step 1) runs scripts/check_update.py which fetches a remote SKILL.md from raw.githubusercontent.com (a public GitHub URL) and then instructs the agent to apply updates and re-read the updated SKILL.md, meaning untrusted third‑party content on GitHub is fetched and can change agent behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime update checker (scripts/check_update.py) fetches remote metadata from URLs like https://raw.githubusercontent.com/{owner}/{repo}/main/.agents/skills/{skill_name}/SKILL.md and uses the remote version to force an immediate update (via the printed "npx skills update ..." command), which can cause the agent operator to fetch and run remote code—so the fetched URL is used at runtime to control behavior that leads to code execution.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill mandates running a local update script and executing an arbitrary UPDATE_COMMAND (automatically attempting updates), which requires executing code that can modify files or the system state even if it doesn't explicitly request sudo, so it pushes the agent towards potentially state-changing and unsafe actions.