The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as test fixtures, protocol replies, and CI logs which could potentially contain malicious instructions meant to influence the agent's behavior.
Ingestion points: Test fixtures, protocol replies, named contracts, and CI logs (SKILL.md, references/measurement-and-ci.md).
Boundary markers: No explicit instructions for delimiters or warnings to ignore embedded instructions in data were found.
Capability inventory: The skill utilizes subprocess execution for running tests and performs file operations for audit logging (SKILL.md).
Sanitization: No specific sanitization or filtering logic for external content is mentioned.
[COMMAND_EXECUTION]: The skill executes repository-specific commands and relies on the project's pinned toolchain. While these are necessary for CI auditing, this execution model relies on the security of the target repository's environment and configuration.
[DYNAMIC_EXECUTION]: As part of its primary auditing purpose, the skill instructs the agent to perturb code behavior and inject controlled failures into isolated copies of the source to verify existing test protections. This runtime modification and execution of code is a functional requirement but represents a dynamic execution surface.

rea-ci-tests-audit