design-handoff-flutter
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted project documentation, such as
mockup-brief.mdandwireframe.md, which are used to define implementation specifications that the agent must follow and verify. - Ingestion points: Untrusted data is identified in
scripts/load_handoff.pyand subsequently parsed byscripts/verify_handoff.pyfrom the 'Implementation spec' section of the design brief. - Boundary markers: The skill incorporates a specific utility (
scripts/_untrusted_doc.py) to wrap external documentation in<<<UNTRUSTED_PROJECT_DOC>>>tags, and theSKILL.mdfile explicitly instructs the agent to treat these inputs as data only and ignore any embedded instructions or role changes. - Capability inventory: The skill has the capability to read and write files within the project root and execute local scripts for verification and layout synchronization.
- Sanitization: The
scripts/verify_handoff.pyscript employsre.escape()on requirement tokens extracted from untrusted tables to prevent regex injection during the automated code auditing process.
Audit Metadata