plan-video
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill acts as a 'thin orchestrator', meaning it strictly coordinates other internal skills and tools without executing arbitrary logic or content generation on its own.
- [DATA_EXFILTRATION]: File operations are confined to the local
./youtube/directory for managing episode-specific research and planning documents. There are no attempts to access sensitive system files (e.g., SSH keys, env files) or exfiltrate data to external domains. - [COMMAND_EXECUTION]: The skill utilizes a set of predefined YouTube MCP tools and internal skill namespaces (
content-strategy,visual-design,writing). It does not use dangerous functions likeeval()or spawn uncontrolled subprocesses. - [PROMPT_INJECTION]: Instructions within the skill are procedural and focused on maintaining workflow integrity (e.g., ensuring research is loaded before planning). There are no instructions aimed at bypassing AI safety filters or overriding core agent behavior.
- [EXTERNAL_DOWNLOADS]: All references to external patterns (e.g.,
youtube-title-formulas.md) are local files within the skill's reference directory. No remote code or scripts are downloaded from untrusted sources.
Audit Metadata