bilibili-subtitle-downloader

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

• Third-party content exposure detected (high risk: 0.90). This skill's scripts (scripts/download_and_chunk.py and scripts/cheese_downloader.py) fetch subtitles from public Bilibili APIs and subtitle URLs and SKILL.md instructs the agent/sub-agent to read those downloaded subtitle chunk files to produce summaries, so untrusted, user-generated third‑party content is ingested and can influence agent behavior.

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

• Potentially malicious external URL detected (high risk: 0.90). At runtime the skill fetches subtitle JSON from Bilibili endpoints (e.g. https://api.bilibili.com/x/player/v2, https://api.bilibili.com/x/web-interface/view, https://api.bilibili.com/pugv/view/web/episode and the dynamic subtitle_url returned) and writes those external subtitles into chunk files that are then fed verbatim into the sub-agent/LLM prompts, so external URLs directly control the model input.