codex-marketing-skills
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, specifically YouTube transcripts (via Supadata API) and sponsorship email content (via Gmail API). * Ingestion points: Untrusted data enters the agent context through the get_transcript and fetch_brand_deal_emails functions in SKILL.md. * Boundary markers: The implementation lacks delimiters or instructions for the agent to ignore embedded instructions within the fetched content. * Capability inventory: The skill possesses capabilities for subprocess execution, file writing, and network communication. * Sanitization: There is no evidence of sanitization or filtering of external content.
- [COMMAND_EXECUTION]: The skill relies on several external CLI utilities (readwise, excalidraw-cli, genmedia) executed through subprocess.run. The Python wrappers interpolate user-controlled variables directly into shell command arguments, creating a surface for command injection.
- [EXTERNAL_DOWNLOADS]: The skill requires cloning a repository from an untrusted source (github.com/rbrown101010/codex-marketing-skills) and installing multiple external Node.js and Python dependencies that are not managed via standard versioned configuration files within the skill itself.
Audit Metadata