run-ui-registry-variations
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a bash script
scripts/validate-registry-access.shwhich executescurlto validate registry connectivity. It also suggests usingsops-secret-accessandcurldirectly for manual validation. While scoped to registry validation, these are active shell commands. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external UI component registries (
shadcnblocks,Aceternity, andshadcn.darkmatter.io) to build UI variations. Malicious content in a registry (e.g., a component's name or metadata containing instructions) could influence the agent's behavior during the build process. - Ingestion points: Fetches from
shadcnblocks,Aceternity, andshadcn.darkmatter.ioregistries. - Boundary markers: None explicitly defined for the external component data.
- Capability inventory: Executes shell scripts (
validate-registry-access.sh), usescurl, potentially usessops, and writes multiple local files for UI variations. - Sanitization: The skill instructions emphasize non-secret evidence in reports but lack specific sanitization for registry-returned component data.
Audit Metadata