validating-and-publishing-canvases

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the retrieval and modification of canvas source projects which include source code and assets (project.files). This creates a surface where the agent processes external data that could contain instructions or malicious code meant to influence the agent's behavior or exploit the canvas environment.
  • Ingestion points: Source code and configuration are ingested via canvas-source-retrieve and user-provided project files.
  • Boundary markers: The platform uses explicit capabilities declarations and a server-side validation service (canvas-validate-create) to identify and block unauthorized operations.
  • Capability inventory: The skill uses tools to read project state, validate code, publish new versions, and manage drafts (canvas-publish-create, canvas-edit-create, canvas-promote-create).
  • Sanitization: Validation diagnostics specifically check for forbidden patterns like forbidden_dynamic_import, forbidden_require, and undeclared network origins.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 10:55 AM
Security Audit — agent-trust-hub — validating-and-publishing-canvases