The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill parses content and filenames from the local prds/ directory to drive its workflow. This establishes a surface area where malicious instructions embedded in a PRD file could attempt to influence agent operations.
Ingestion points: The skill reads PRD files located at prds/[issue-id]-[feature-name].md to validate readiness and extract task information (Step 1).
Boundary markers: The instructions do not specify the use of delimiters or "ignore" instructions to separate untrusted PRD content from the agent's logic.
Capability inventory: The skill utilizes git for branch management and forge CLIs like gh, glab, or tea for issue tracking (Step 3).
Sanitization: No explicit data sanitization or validation of the extracted issue IDs or feature names is implemented before they are interpolated into shell commands.

prd-start