process-hybrid-search

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses pnpm dlx to download and execute the official @tiangong-lca/cli package (version 0.1.16) from the public NPM registry. This is consistent with the skill's stated purpose of providing a zero-config wrapper for the vendor's search tool.\n- [COMMAND_EXECUTION]: Shell commands are executed via node:child_process to verify the environment (pnpm version) and run the search CLI. The commands are constructed using static strings and validated arguments.\n- [DYNAMIC_EXECUTION]: The skill supports a local development mode via the --cli-dir flag, which triggers an automated pnpm install and pnpm build workflow. This process is protected by strict manifest verification, including SHA256 checksum validation for all schema files and identity checks for the local package.json.\n- [INDIRECT_PROMPT_INJECTION]: As a search-oriented skill, it processes user-supplied queries that are sent to a remote Supabase Edge Function. While this presents a theoretical surface for indirect injection if search results contain malicious content, the skill includes explicit instructions for the AI to preserve constraints and avoid credential handling.\n
  • Ingestion points: User queries provided via scripts/run-process-hybrid-search.mjs or assets/example-request.json.\n
  • Boundary markers: The references/prompts.md file defines strict requirements for the AI when interpreting inputs.\n
  • Capability inventory: Execution is limited to the search CLI and associated build tools.\n
  • Sanitization: The skill delegates authentication and request construction to the official CLI tool, which is responsible for safe data handling.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 04:28 AM