mobile-first-website-design
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill exhibits several robust safety and integrity features:
- [INDIRECT_PROMPT_INJECTION]: While the skill researches external design patterns, it mitigates risk through its primary workflow design:
- Ingestion points: Research enters the context from external tools like Refero and Mobbin (SKILL.md).
- Boundary markers: The workflow enforces 'freezing' wireframes with SHA-256 hashes before identity work begins, ensuring integrity (SKILL.md).
- Capability inventory: The skill uses controlled tools like
browser-visual-qaand local python scripts (scripts/validate_packet.py) for validation. - Sanitization: Instructions mandate normalizing all inputs, audience data, and research into canonical JSON at the start of the contract (SKILL.md).
- [DYNAMIC_EXECUTION]: Local script execution is restricted to internal integrity checks and validation rules (
scripts/prompt_corpus.py,scripts/breakpoint_rules.py). - [DATA_EXFILTRATION]: Network operations are confined to research tools and a specified MCP endpoint (
https://agents.flora.ai/mcp), with explicit instructions never to store secrets or fabricate results. - [COMMAND_EXECUTION]: Command patterns found (e.g.,
mise run ci) are standard CI/CD workflow markers and do not involve unsanitized user input.
Audit Metadata