claude-design-system-hooks
Warn
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches core components and project files from a personal GitHub repository (
BharathKumarSuresh/claude-design-system-hooks). This repository does not appear to be an official organizational source for the claimed service. - [REMOTE_CODE_EXECUTION]: Executes code via the installation of Node.js packages from the unverified
@claude-designscope and community-provided plugins from a marketplace under the@communityscope. - The skill instructs the user to pull a Docker image (
claude-design-engine:2026-stable) which is not hosted on a verified or official registry. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external design specifications, Figma exports, and image mockups to generate production-ready source code, which could be exploited to generate malicious code if the input sources are compromised.
- Ingestion points: Processes external files such as
design-system-spec.yaml,figma-export.json, and image assets through the CLI. - Boundary markers: There is no mention of boundary markers or instructions to the AI engine to ignore instructions embedded within design specifications.
- Capability inventory: The
claude-designCLI tool has direct capabilities to write generated code, tests, and documentation to the local filesystem (e.g., using--output ./src/components). - Sanitization: The instructions do not specify any validation or sanitization protocols for the design specifications or external data being processed.
Audit Metadata