setup
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external, user-provided content such as websites and images to gather design references. This creates a surface where malicious instructions embedded in those external resources could potentially influence the agent's subsequent actions.
- Ingestion points: Links or images provided by the user in
SKILL.md(Step 3) andreferences/interview.md(Step 5). - Boundary markers: Absent; the instructions do not specify delimiters or warnings to ignore instructions found within the external reference material.
- Capability inventory: The agent is authorized to write files to the project root (
DESIGN.md), execute shell commands for rendering, and invoke other functional skills likecraft:systemandcraft:sync. - Sanitization: Absent; no mention of sanitizing or filtering content fetched from the provided URLs.
- [COMMAND_EXECUTION]: The skill defines specific shell command patterns to be used for rendering and capturing screenshots of local HTML files.
- Evidence:
references/adapters/html.mdprovides a specific command template for headless Chrome:"<chrome>" --headless=new --screenshot=references.png --window-size=1280,900 file://$PWD/references.html.
Audit Metadata