design-taste-frontend
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill contains a curl command in SKILL.md to fetch content from a pinned commit in the Leonxlnx/taste-skill GitHub repository for provenance verification. This operation is restricted to reading the file header and does not execute remote code.
- [COMMAND_EXECUTION]: The manifest specifies example commands for installation and validation using npx, node, and bash. These are standard developer tools used for managing the skill lifecycle and point to the author's repository.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided design briefs, which is a potential surface for indirect injection. This is mitigated by a mandatory design read declaration and a structured pre-flight checklist. Ingestion points include user prompts, while the pre-flight check serves as a security gate. Capability inventory includes access to Bash and file system tools as defined in the YAML frontmatter.
- [DYNAMIC_EXECUTION]: A local Node.js script is included for structural validation of the skill's evaluation suite. The script uses standard built-in modules for static analysis of JSON data and does not utilize unsafe evaluation methods.
Audit Metadata