The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided session data and human inputs, which are then summarized and written back to external ticket systems as HTML comments.- Ingestion points: The cawplan session collect command in SKILL.md reads user-provided human inputs from local session data into the agent's context.- Boundary markers: The skill instructs the agent to summarize inputs rather than performing raw copies, providing a layer of abstraction between the input and the final output.- Capability inventory: The cawplan tickets update command in SKILL.md allows the agent to perform network operations to write data back to a cloud-based ticket system.- Sanitization: The instructions explicitly require the agent to escape session text before placing it into the generated HTML progress_comment.

cawplan-coding-commit