yt-community-post-writer
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources, specifically YouTube video URLs (processed via
lib/url_parser.py) and user-provided samples of past writing to build a voice profile. This represents a potential surface for indirect prompt injection. The risk is mitigated by the design of the skill, which requires a human-in-the-loop approval step before any content is finalized or 'published' (which the skill describes as returning a copy-paste block to the user). Evidence of this mitigation is found inSKILL.mdstep 5 andsub-skills/illustration.mdstep 5. - [COMMAND_EXECUTION]: The skill relies on internal library functions (
lib.publish,lib.illustrate,lib.refine) to perform its core tasks. These functions are vendor-provided wrappers used for formatting text and interfacing with image generation APIs. The skill's description confirms that even the 'publish' function only returns text for the user to manually handle, ensuring no unauthorized external actions occur.
Audit Metadata