cross-niche-outliers

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly scrapes and ingests user-generated content from public sources (YouTube via yt-dlp and youtube-transcript-api/Apify in scripts/scrape_cross_niche_outliers.py and scrape_cross_niche_tubelab.py, and TubeLab API in scrape_cross_niche_tubelab.py) and then has the agent summarize and act on those transcripts/titles (feeding them into summarization and title-generation steps), so third-party content can materially influence tool decisions and outputs.

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

• Potentially malicious external URL detected (high risk: 0.70). The scripts call external services at runtime—notably the TubeLab API (https://public-api.tubelab.net/v1) and the Apify "karamelo/youtube-transcripts" actor (via APIFY_API_TOKEN)—and the returned raw transcripts/metadata are injected into prompts sent to Anthropic/Claude for summarization and title generation, so externally fetched content directly enters and can control the model context.