runway
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes established, official command-line interfaces (CLIs) for all deployment actions, ensuring that operations are performed within the security boundaries of the target platforms. It explicitly instructs the agent to avoid printing or echoing sensitive environment variable values.\n- [INDIRECT_PROMPT_INJECTION]: The skill incorporates an attack surface for indirect prompt injection by processing external project configurations and deployment receipts, though it includes mitigations.\n
- Ingestion points: Reads project configuration files (e.g., vercel.json, fly.toml) and a JSON-formatted receipt file via the
check-receipt.pyscript.\n - Boundary markers: The skill requires explicit user authorization for critical production actions and mandates the verification of immutable version fingerprints to ensure the live behavior matches the intended build.\n
- Capability inventory: Includes execution of platform CLIs and a bundled Python script for data validation.\n
- Sanitization: The
scripts/check-receipt.pyutility provides automated validation of receipt data, checking for correct types, mandatory fields, and consistent build identifiers before accepting the evidence.
Audit Metadata