yt-script
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill invokes a local Python utility,
hookscore.py, to analyze and rank YouTube hook candidates. - Evidence: Shell commands in
SKILL.mddemonstrate the agent runningpython3 hookscore.py hooks.txtto process local files. - Analysis: The script logic is focused on text evaluation using regular expressions and standard library modules. It does not perform network requests, spawn shells, or use dangerous dynamic execution functions.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (user-provided video transcripts) to infer and persist a 'voice profile' used for future script generation.
- Ingestion points: User-supplied videos or transcripts provided during the voice inference step in
SKILL.md. - Boundary markers: There are no explicit instructions or delimiters defined to separate the untrusted video content from the agent's logic for inferring the voice profile.
- Capability inventory: The skill has the ability to write to the file system (
~/.claude/youtube/voice.md) and execute a local Python script. - Sanitization: No filtering or validation steps are included to sanitize the inferred voice profile before it is stored or reused to guide the agent's writing behavior.
Audit Metadata