skills-sync

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

• Third-party content exposure detected (high risk: 0.90). The skill's install flow in scripts/sync.py (cmd_install) explicitly clones arbitrary git URLs (http/https/git@) into the local ~/.ai-skills/project skill directories and then syncs/copies those untrusted, user-provided repositories into IDE skill targets, meaning third-party, user-generated content is fetched and incorporated into the agent's runtime skill set.

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

• Potentially malicious external URL detected (high risk: 0.80). The script's install command will, at runtime, run git clone on arbitrary sources that start with "http://", "https://" or "git@" (subprocess.run(["git","clone", source, ...])), which fetches remote repositories that contain SKILL.md files that are then copied/synced into IDE skill directories and can directly control agent prompts (flagged pattern: git clone of http(s)://... or git@github.com:...).