youtube-analyzer

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and ingests public third‑party content — e.g., Phase 1 uses yt-dlp and youtube_transcript_api to download YouTube metadata/transcripts (references/source-selection.md, Phase 1–Step 1.2/1.3) and Phase 3 clones and has Explore agents read GitHub repos (references/scaling-and-repo-explore.md / workflows/repo-exploration.md), and those external transcripts and repoExploreResults are directly read and used by workflow and synthesis agents to drive analysis, package DB updates, and follow-up actions, so untrusted user-generated content can materially influence tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly runs git clone at runtime ("git clone --depth 1 "{repoUrl}" "{scratchpad}/repo-explore/""), fetching arbitrary GitHub repositories which are then parsed by Explore sub-agents and injected into synthesis prompts—i.e., external repo content is fetched during runtime and directly controls agent inputs.