thoughtproof-reasoning-check

Pass

Audited by Gen Agent Trust Hub on Mar 27, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill transmits reasoning data (claims) to an external endpoint https://api.thoughtproof.ai/v1/check. This involves sending financial intent and context to an external third-party domain not explicitly identified as a vendor resource.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) via the claim field. Maliciously crafted claims could include instructions designed to influence the verdict of the multi-model critique to return an ALLOW status for risky actions.\n
  • Ingestion points: claim field in the JSON payload of the POST request (SKILL.md).\n
  • Boundary markers: Absent; the claim content is interpolated without delimiters or safety instructions.\n
  • Capability inventory: The verification output (ALLOW/BLOCK) is designed to control high-value transactions in secondary skills such as moonpay-swap-tokens.\n
  • Sanitization: Absent; the skill does not specify validation or filtering for the processed claim text.\n- [COMMAND_EXECUTION]: The skill documentation provides examples using curl to interact with the remote API, which involves shell-level command execution for network communication.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 27, 2026, 03:34 AM