The Agent Skills Directory

[SAFE]: No malicious code, obfuscation, or unauthorized access attempts were identified. The skill's behavior is consistent with its stated purpose of design refinement.
[COMMAND_EXECUTION]: The skill uses the WebSearch tool to look up design references for products listed in the brand direction file. This is a legitimate tool use for the design workflow.
[SAFE]: The skill includes strong security controls, specifically a human-in-the-loop requirement ([STOP — BLOCKING]) that prevents any file updates until the user has reviewed and confirmed the proposed changes.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests data from local documentation and web search results without explicit boundary markers. However, the risk is mitigated by the disable-model-invocation: true configuration and the mandatory user approval gate.
Ingestion points: docs/product/design/brand-direction.md, docs/product/design-principles.md, docs/product/personas/, and WebSearch results (SKILL.md)
Boundary markers: Absent
Capability inventory: WebSearch and file write to docs/product/design/brand-direction.md (SKILL.md)
Sanitization: Absent

recipe-refine-visuals