agent-canvas-environment

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses local shell commands including curl and jq to interact with the Agent Canvas API at http://localhost:8001. It performs tasks such as session validation, settings retrieval, and conversation creation.
  • [DATA_EXFILTRATION]: The skill accesses sensitive information by reading the Agent Canvas session API key from environment variables or the local file path ~/.openhands/agent-canvas/api-key.txt. This data is used solely for authenticating requests to the local backend service.
  • [PROMPT_INJECTION]: The skill implements a delegation pattern that creates new agent conversations using POST /api/conversations. It configures these sub-agents with increased autonomy, such as setting the confirmation_policy to NeverConfirm and increasing max_iterations, which creates a surface for indirect instructions to execute without user intervention.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 06:21 PM
Security Audit — agent-trust-hub — agent-canvas-environment