validating-and-publishing-canvases
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and iterate upon external source projects which introduces a risk of indirect prompt injection if those files contain malicious instructions.
- Ingestion points: Project files and configuration metadata are retrieved into the agent's context through the canvas-source-retrieve tool as described in SKILL.md.
- Boundary markers: The instructions do not define specific delimiters or security prompts to help the agent distinguish between the canvas source code and the operational publishing instructions.
- Capability inventory: The agent is granted tools to read, rewrite, edit (canvas-edit-create), and promote (canvas-promote-create) code to the PostHog platform.
- Sanitization: While the platform performs technical validation for specific constraints like allowed HTTPS origins and dependency versions, it does not provide mechanisms to sanitize or filter natural language prompt instructions embedded in the project files.
Audit Metadata