The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided content (brain dumps, notes, and research) to generate various content formats, which creates an indirect prompt injection surface.
Ingestion points: Data enters the agent's context through placeholders in the prompt files: {notes} in prompts/first-draft.md, {text} in prompts/clarity-pass.md, prompts/empathy-rewrite.md, and prompts/polish-pass.md, and {source_content} in prompts/remix-engine.md.
Boundary markers: The templates use markdown headers and block delimiters (e.g., [NOTES], [SOURCE]) to separate user input from the agent's instructions, reducing the risk of accidental instruction confusion.
Capability inventory: The skill utilizes a platform-specific sessions_spawn command to create new agent sessions for specific tasks. It does not perform network operations or file system modifications.
Sanitization: The skill relies on formatting and delimitation but does not implement programmatic escaping or validation of the interpolated content.
[COMMAND_EXECUTION]: The skill instructs the agent to utilize a platform-specific command to manage parallel content production tasks.
Evidence: SKILL.md contains instructions for the agent to use sessions_spawn --task "..." to delegate writing and remixing tasks to sub-agent sessions.

content-factory