skills-cli

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md workflow and references/cli.md explicitly let the agent fetch and install skills from GitHub URLs and other git remotes via bunx skills add <source> (and the Guardrails note that installed skills are "executable instructions"), so the agent can ingest untrusted third‑party repositories that can materially change its behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill instructs the runtime command "bunx skills add https://github.com/sjunepark/custom-skills/tree/main/skills" (and other git remotes like vercel-labs/skills) to fetch remote skill definitions which are installed as executable instructions that directly control agent prompts/behavior, so remote content fetched at runtime can control the agent.