agents-md
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The
references/evals.yamlfile contains simulated prompt injection strings used strictly for testing and evaluating the agent's security resilience; these do not represent malicious instructions for the agent's operation. - [DATA_EXFILTRATION]: The skill instructions include explicit constraints to redact secrets, tokens, credentials, and PII from all generated documentation outputs.
- [COMMAND_EXECUTION]: The skill guides the agent to use repository-native tools (e.g.,
npm,git,python3) and existing scripts (e.g.,codex-preflight.sh) for environment discovery and validation. - [PROMPT_INJECTION]: The skill processes untrusted instruction files from repositories, representing an indirect prompt injection surface. Ingestion points: Reads existing
AGENTS.mdand linked markdown documents. Boundary markers: None explicitly enforced for the ingestion of untrusted repository content. Capability inventory: Execution of verified repository-native commands and scripts. Sanitization: Mandatory redaction of PII and sensitive credentials from documentation outputs.
Audit Metadata