The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it ingests untrusted data from multiple external project files to initialize or sync the board state.
Ingestion points: The board init and board sync commands read from TODO*, ROADMAP*, PLAN*, BACKLOG*, README files, and the current conversation history (SKILL.md).
Boundary markers: The instructions lack guidance on using boundary markers (e.g., XML tags) or specific "ignore" instructions when the agent parses content from these untrusted external files.
Capability inventory: The skill has the capability to write to the file system through init, sync, and archive operations.
Sanitization: No explicit sanitization or validation of the ingested content is described before it is rendered back to the user or written to disk.
[DATA_EXPOSURE]: The skill uses glob and grep patterns to scan the project directory for various documentation files. While this is intended for functionality, it involves the agent processing arbitrary file content from the workspace.

board