graffiti-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch documentation, templates, and system primitives from the official Graffiti UI website (https://graffiti-ui.com). These references are consistent with the vendor infrastructure and serve as the authoritative source for validating framework compliance.
- [INDIRECT_PROMPT_INJECTION]: The skill handles user-supplied design requirements and existing code snippets for refactoring. It mitigates potential injection risks by enforcing a strict 'Output Contract' and an 'Anti-Patterns' catalog that prevents the agent from executing arbitrary instructions hidden in the input.
- Ingestion points: User-provided prompts and code snippets for conversion or refactoring as specified in SKILL.md.
- Boundary markers: Mandatory quality gates in references/OUTPUT_CONTRACT.md and refusal rules in SKILL.md define clear operational limits for the agent.
- Capability inventory: The skill is limited to generating static HTML and Svelte markup; it does not possess capabilities for shell execution, persistence, or arbitrary file system access.
- Sanitization: The skill mandates strict validation of all generated classes and tokens against the official CSS contracts defined in the grounding documentation.
Audit Metadata