triage-finding

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and reads untrusted, public third-party content — e.g., "Website/article link — fetch with WebFetch", GitHub links via gh/WebFetch, YouTube transcripts via Exa and youtube-transcript-api, and Telegram/posts/screenshots — as described in SKILL.md and README.md, and uses that content to fact-check and drive triage recommendations, so external content can influence agent decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches user-supplied links at runtime (e.g., https://github.com/some/cool-tool) using WebFetch/gh and transcript-fetching tools and injects that fetched content into the agent's context to drive its triage prompts, so remote content can directly control instructions.