add-adapter
Pass
Audited by Gen Agent Trust Hub on Jul 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [Indirect Prompt Injection Surface]: The skill is designed to ingest data from external GitHub repositories and README files provided by users. This creates a potential surface where untrusted data could influence the agent's behavior during the generation of project files.\n
- Ingestion points: External GitHub repository URLs and documentation/README files.\n
- Boundary markers: The skill lacks explicit prompt delimiters for external data, relying on instructions to 'never invent details.'\n
- Capability inventory: The agent can modify TypeScript source files and execute shell commands (pnpm build, pnpm test).\n
- Sanitization: No specific sanitization of external text is mentioned before interpolation into project files.\n- [Dynamic Code Modification and Execution]: The skill modifies project source code and configuration files based on information extracted from external sources and subsequently triggers validation processes. This involves updating files like 'integration-tests' and 'scaffold-spec.ts' and then executing 'pnpm build' and 'pnpm test'.\n- [Shell Command Execution]: The skill uses local shell commands, including 'ls' for directory listing and 'pnpm' for building, testing, and type-checking the repository to ensure registry parity and integrity.\n- [External Content Reference]: The workflow involves fetching information from external GitHub repositories. While these are well-known services, the ingestion of their content is a standard part of the adapter registration process.
Audit Metadata