styleseed
Warn
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions facilitate the installation of code from an external, unverified GitHub repository (
bitjaru/styleseed) vianpx skills addand the Claude plugin marketplace. - [REMOTE_CODE_EXECUTION]: The skill documentation explicitly states that it runs bundled Node.js scripts for tasks like 'context compilation' and 'quality gates'. These scripts are part of the external package downloaded from GitHub, resulting in the execution of unverified remote code.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process a 'design lock' file (
STYLESEED.md) and a.styleseed/directory to derive design rules. This creates an attack surface where an attacker (e.g., via a malicious Pull Request) could inject instructions into these files to influence the agent's behavior. - Ingestion points:
STYLESEED.mdand.styleseed/directory content. - Boundary markers: No delimiters or instructions to ignore embedded content were found in the provided files.
- Capability inventory: The skill performs Node.js script execution and writes project-specific files.
- Sanitization: No evidence of sanitization or content validation for the input files was observed.
Audit Metadata