The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and analyze external, potentially untrusted data such as source code and test suites, which creates an indirect prompt injection attack surface. This allows for the possibility that malicious instructions embedded in the code being reviewed could influence the agent's behavior.
Ingestion points: source-code, test-suite, and intake-brief are defined as primary inputs in the skill metadata.
Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions when reading these external files.
Capability inventory: While the skill itself is prompt-based and ships no scripts, it is intended for use by agents that typically possess file-system access to read and report on the provided source materials.
Sanitization: No sanitization or validation steps are mentioned for the external content before it is processed by the agent.

implementation-integrity-audit