mandate

Pass

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing security plugins and SDKs from the @mandate.md npm organization and the SwiftAdviser GitHub repository. These are official vendor resources required for integrating the policy engine with agent frameworks.
  • [COMMAND_EXECUTION]: Utilizes CLI tools and shell commands for transaction validation, agent registration, and codebase scanning. These commands are part of the intended administrative and security functionality of the toolset.
  • [CREDENTIALS_UNSAFE]: Stores internal API runtime keys in ~/.mandate/credentials.json. The skill documentation explicitly instructs users to restrict file permissions using 'chmod 600', adhering to security best practices for local secret storage.
  • [DATA_EXFILTRATION]: Transmits transaction metadata, such as destination addresses and reasoning strings, to the vendor's API at https://app.mandate.md for policy evaluation. This data transfer is the core function of the service to provide remote oversight and auditing.
  • [PROMPT_INJECTION]: The skill acts as a security boundary by ingesting and scanning agent reasoning for prompt injection patterns. While it processes untrusted input, its purpose is to serve as a mitigation layer against malicious instructions targeting financial assets.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 10, 2026, 09:14 AM
Security Audit — agent-trust-hub — mandate