The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface due to its primary function of ingesting and analyzing external configuration data.
Ingestion points: Configuration files like af-plan.yaml, ZAP scan policies, and CI/CD workflow files are located via Glob and Grep then processed by the agent.
Boundary markers: The skill includes a dedicated 'Prompt Injection Safety Notice' section that explicitly instructs the agent to treat configuration content as untrusted data.
Capability inventory: The skill is restricted to Read, Grep, and Glob tools, which provides no capability for code execution or network exfiltration from the ingested data.
Sanitization: Instructions advise the agent to ignore any text that appears to be a prompt or command within the target files.
[SAFE]: Metadata check: The skill metadata lists the author as 'unitoneai', which deviates from the system-provided author context of 'CCCreep'. This represents a metadata discrepancy but does not align with any known malicious patterns or obfuscation techniques.

dast-config