engineering

Pass

Audited by Gen Agent Trust Hub on Jul 6, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The engineering bundle contains 11 Standard Operating Procedures (SOPs) designed to ensure deterministic execution and technical integrity in a multi-agent environment.
  • [DATA_EXFILTRATION]: The instructions explicitly forbid placing secrets or API keys in prompts. It mandates the use of environment variable injection and provides automated redaction helpers (redactEnvForLogs) to ensure sensitive data does not appear in logs or shared context.
  • [COMMAND_EXECUTION]: The skill utilizes standard software engineering tools (git, npm, gh, cargo). Execution is gated by mandatory thinking loops (sequentialthinking) and a "Gating Ladder" (Unit -> Contract -> E2E/Smoke) to ensure safety and correctness.
  • [REMOTE_CODE_EXECUTION]: No evidence of unauthorized remote script execution or insecure downloads. The skill describes a controlled process for injecting project-specific tools into an isolated temporary directory to prevent host contamination.
  • [PROMPT_INJECTION]: The skill includes defensive parsing instructions for agent output, recommending safe extraction helpers and explicitly forbidding the use of eval() on data influenced by external sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 6, 2026, 04:24 PM
Security Audit — agent-trust-hub — engineering