The Agent Skills Directory

[SAFE]: The skill performs legitimate file system operations, reading from and writing to project-specific paths (e.g., ./youtube/episode/.../research.md and plan.md). These operations are consistent with its stated purpose of content planning.
[SAFE]: The skill maintains user oversight by explicitly requiring user selection for titles, thumbnails, and hooks before proceeding with the planning workflow.
[SAFE]: Indirect Prompt Injection Surface Analysis:
Ingestion points: Reads research data from ./youtube/episode/[episode_number]_[topic_short_name]/research.md.
Boundary markers: None explicitly defined in the prompt instructions for the ingested data.
Capability inventory: Performs file writes to plan.md and invokes other platform skills (youtube-title, thumbkit, youtube-thumbnail).
Sanitization: No explicit sanitization or validation of the research file content is performed. While this represents a standard data ingestion surface, there is no evidence of malicious exploitation or instructions that bypass safety protocols.

youtube-content