The Agent Skills Directory

[SAFE]: The skill implements a strict read-only policy for linting checks, explicitly forbidding the modification of source files, ignore settings, or rule thresholds.
[COMMAND_EXECUTION]: The skill is configured to run local linting binaries like codeguard. These operations are scoped to project-specific paths and require the agent to disclose any potential network access to the user before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository files and command-line outputs from linting tools (Ingestion Points). It uses structured output templates to isolate results (Boundary Markers) and limits its capabilities primarily to local read-only linter execution (Capability Inventory). The skill instructions explicitly require the desensitization of credentials, authentication tokens, and private paths in the final report (Sanitization).

codeguard-check