The Agent Skills Directory

[PROMPT_INJECTION]: The skill is subject to indirect prompt injection because it ingests untrusted data from the repository (e.g., Platform-Engineering-Constitution.md and repository artifacts) to derive audit rules and findings. Ingestion points: Platform-Engineering-Constitution.md and various source code files (IaC, Dockerfiles, application code). Boundary markers: Absent. Capability inventory: Limited to read-only search operations (find, grep). Sanitization: Absent. Mitigation recommendations include wrapping external content in delimiters with explicit instructions to ignore embedded commands and using structured data validation for audit rules.
[COMMAND_EXECUTION]: The skill utilizes standard system utilities (find, grep) to search the local filesystem for infrastructure artifacts and patterns like potential secret exposures. This behavior is consistent with its stated purpose as an auditing tool and is executed within a read-only guardrail.

app-audit