content-ideator
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted content from social media captions, transcripts, and page text, creating a surface for indirect prompt injection attacks. Ingestion points: External content scraped from social media platforms via yt-dlp and Apify. Boundary markers: The skill contains an explicit warning in SKILL.md instructing the agent to treat scraped text as data only and to stop if addressed directly. Capability inventory: Subprocess execution (ffmpeg, yt-dlp), Airtable writes, Supabase reads, and local file system writes. Sanitization: No programmatic sanitization or strict schema validation is described for the scraped content.
- [COMMAND_EXECUTION]: The skill utilizes shell commands including
yt-dlpandffmpegvia subprocess calls to download and process video content for transcription and analysis. - [EXTERNAL_DOWNLOADS]: The skill fetches external content from social media platforms using
yt-dlpand an Apify actor (shu8hvrXbJbY3Eb9W) to retrieve media and metadata.
Audit Metadata