shadcn-registry-first
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to run local setup scripts (
scripts/setup-components-json.sh) and utilize thebunxpackage runner to execute theshadcnCLI for adding components. - [INDIRECT_PROMPT_INJECTION]: The agent ingests component names, descriptions, and code from the
@shadcnand@shadcnblocksregistries. This creates a potential surface for instructions embedded in component metadata to influence the agent's logic during the shortlisting or installation process. - Ingestion points: Tool output from
search_items_in_registriesandget_item_examples_from_registries. - Boundary markers: The instructions do not define explicit delimiters or "ignore embedded instructions" warnings for the registry-sourced data.
- Capability inventory: The agent has the ability to write files to the project directory and execute shell commands via
bunx. - Sanitization: No specific sanitization or filtering logic is provided for strings retrieved from the registries.
Audit Metadata