The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions establish a surface for indirect prompt injection by ingesting untrusted raw material and re-reading the generated article file from disk. \n- Ingestion points: The markdown file provided by the user and the article file saved to disk (SKILL.md). \n- Boundary markers: Absent; there are no specific markers or instructions for the agent to differentiate between text data and potential instructions. \n- Capability inventory: The skill performs file read and write operations on the local system (SKILL.md). \n- Sanitization: Absent; the skill does not specify any sanitization or validation of the input material before it is used in the narrative or written to the file system.

writing-beats