The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to implement features based on external GitHub issues and their associated comments, creating an attack surface where malicious users could provide directives to override the agent's intended behavior.
Ingestion points: The skill ingests untrusted external data using the command gh issue view <n> --comments and by reading parent specifications.
Boundary markers: No delimiters or safety instructions are defined to separate implementation specifications from potentially malicious instructions within comments.
Capability inventory: The skill possesses significant capabilities, including the execution of shell commands for testing/typechecking and interacting with GitHub via gh pr create and gh pr comment.
Sanitization: No sanitization or validation logic is present to filter or verify the integrity of the instructions found in issue comments.
[COMMAND_EXECUTION]: The skill invokes local command-line tools including the GitHub CLI (gh) and project-specific test suites or typecheckers. This functional requirement provides the mechanism through which an indirect injection attack could execute malicious actions on the host environment.

implement