reference-first
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources including GitHub repositories, package registries, and third-party documentation URLs discovered during search tasks.\n
- Ingestion points: Data enters the agent's context through GitHub search results (READMEs, code snippets), package metadata, and external technical documentation.\n
- Boundary markers: The protocol does not explicitly require strict delimiters for external content, though it emphasizes 'synthesising' rather than direct quoting.\n
- Capability inventory: The skill can search GitHub, recommend new project dependencies, and formulate design blueprints based on discovered patterns.\n
- Sanitization: The instructions include a safeguard requiring the agent to manually rewrite patterns in the project's local convention instead of pasting source files directly.\n- [COMMAND_EXECUTION]: Executes GitHub CLI commands (
gh search repos,gh search code) and package registry queries (npm view) to gather technical references and metadata.\n- [EXTERNAL_DOWNLOADS]: Directs the agent to access and analyze documentation and source code from diverse external domains and repositories to inform architectural decisions.
Audit Metadata