The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from X, Instagram, YouTube, and TikTok, which could contain malicious instructions designed to influence the agent's behavior during the aggregation and planning phase.\n- Ingestion points: Data is ingested via the outliers.json and video-analysis.json files produced by subagents (located in x-research/, instagram-research/, youtube-research/, and tiktok-research/ directories), which contain raw text and metadata from social media posts.\n- Boundary markers: The workflow lacks explicit delimiters or instructions for the agent to treat external content as data rather than instructions when generating playbooks and content ideas.\n- Capability inventory: The skill has permission to execute shell commands for directory management and perform comprehensive file writing operations to the project directory.\n- Sanitization: There is no mention of sanitization or validation steps to filter out potentially harmful content from the social media data before it is included in the final output files (content-ideas.md and platform playbooks).

content-planner