The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from issue tracker titles, bodies, and comments, which could contain instructions designed to influence agent behavior.
Ingestion points: Issue tracker descriptions, resolution comments, and map metadata.
Boundary markers: No delimiters or explicit instructions to isolate tracker-sourced data from internal logic are specified.
Capability inventory: The skill can create and modify issues, manage repository labels, and perform Git operations.
Sanitization: Content retrieved from the tracker is not explicitly validated or sanitized.
[COMMAND_EXECUTION]: The skill performs repository-level operations to track project progress and store research results.
Evidence: The instructions mandate creating Git branches (e.g., research/<name>) to capture findings from research tickets.
Evidence: The skill performs 'Wayfinding operations' that involve interacting with issue tracker APIs or local markdown files to update the state of the map.

wayfinder