The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from tickets, session logs, and user feedback which may contain malicious instructions.
Ingestion points: External data enters the agent context through cawplan session collect and cawplan tickets search (SKILL.md).
Boundary markers: The skill explicitly instructs the agent to treat retrieved content as data, not instructions, as defined in the references/UNTRUSTED_CONTENT.md file.
Capability inventory: The skill possesses the capability to perform remote writes via cawplan session report and cawplan tickets update commands (SKILL.md).
Sanitization: The instructions require the agent to escape user-generated content before interpolating it into HTML progress comments for ticket updates (SKILL.md).
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute shell commands for process management, such as launching background tasks and polling PIDs to monitor the state of a local web server (SKILL.md).
[EXTERNAL_DOWNLOADS]: The skill performs network operations to interact with the Cursor Dashboard API at cursor.com and the CawPlan cloud service for data synchronization and reporting tasks (SKILL.md).

cawplan-coding-commit