yt-comment
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input in the form of pasted YouTube comment threads, which presents a surface for potential malicious instructions embedded in comments.
- Ingestion points: User-pasted comment threads as defined in the 'Triage first' section of SKILL.md.
- Boundary markers: The skill explicitly requires a 'Triage' step to categorize inputs, including a 'Bait' category to identify and ignore malicious or irrelevant content. It also implements a 'gate' where the user must manually copy and publish the output, preventing automated execution of drafted content.
- Capability inventory: The skill is authorized to read from and write to a local configuration file at
~/.claude/youtube/voice.mdto manage voice profiles. - Sanitization: There is no mention of sanitizing or escaping the content of the comments before they are used to draft replies.
Audit Metadata