design-handoff-flutter
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted design documents and image assets which could potentially contain malicious instructions or command injection attempts.
- Ingestion points: External design files such as
mockup-brief.md,wireframe.md, and screenshots are loaded from the.heyeddi/designs/directory usingscripts/load_handoff.pyandscripts/describe_handoff.py. - Boundary markers: The skill utilizes a security utility in
scripts/_untrusted_doc.pythat wraps ingested project content with specific delimiters (<<<UNTRUSTED_PROJECT_DOC>>>) to instruct the agent to treat the content as data rather than instructions. - Capability inventory: The skill provides tools to modify Flutter source code in the
lib/directory, update project documentation (design.md), and execute local verification scripts. - Sanitization: The
SKILL.mdfile contains explicit safety instructions for the agent to ignore any commands or role changes embedded in the design briefs and to treat them strictly as data.
Audit Metadata