The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository being analyzed, specifically files such as CODING_STANDARDS.md, docs/standards/*, and .cursor/rules/*. This content is used as the "source of truth" for generating remediation tickets and prioritizing phases.
Ingestion points: CODING_STANDARDS.md, files in docs/standards/, and rules in .cursor/rules/*.mdc.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are implemented for this external data.
Capability inventory: The skill possesses the ability to execute shell commands (npm run), read files, write to the .scratch/ directory, and potentially create GitHub issues.
Sanitization: There is no evidence of sanitization or validation of the text content extracted from these files before it is processed by the model.
[DYNAMIC_EXECUTION]: The skill is designed to invoke local verification scripts defined in the repository's configuration, such as verify, lint, and typecheck (e.g., via npm run). This involves executing code provided by the repository being analyzed, which is a potential risk if the repository source is untrusted.
[COMMAND_EXECUTION]: The instructions require the agent to run command-line tools to gather evidence of standards gaps. It specifically notes how to handle shell execution on Windows (e.g., avoiding pipes that swallow stdio), indicating direct interaction with the host environment's shell.

propose-standards-tickets