consult-chatgpt-pro

Pass

Audited by Gen Agent Trust Hub on May 16, 2026

Risk Level: SAFEDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill is designed to package and transmit local project data to chatgpt.com for review.
  • Exfiltration Channel: Data is sent to chatgpt.com, a well-known service, via the user's desktop browser using Computer Use tools.
  • Sanitization Measures: The skill explicitly instructs the agent to exclude secrets, .env files, API keys, tokens, and unrelated personal data from the transmission packet.
  • User Authorization: Every external submission requires the user to approve the exact packet and verify the destination account/workspace at runtime.
  • [PROMPT_INJECTION]: The skill processes output from an external source (ChatGPT Pro), which introduces a surface for indirect prompt injection attacks.
  • Ingestion Points: External data enters the agent context from the chatgpt.com response captured via the browser.
  • Boundary Markers: The skill utilizes a structured prompt template (review-prompt-template.md) and provides instructions to the external model to treat packet content as data, not instructions.
  • Capability Inventory: The agent using this skill has access to Computer Use, file system operations, and shell execution.
  • Validation & Sanitization: The skill mandates a 'trust but verify' workflow where all external advice is treated as untrusted and must be confirmed against local evidence before any action or recommendation occurs.
Audit Metadata
Risk Level
SAFE
Analyzed
May 16, 2026, 03:36 PM
Security Audit — agent-trust-hub — consult-chatgpt-pro