tiangong-kb-esg-search

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill uses npx @tiangong-ai/cli@0.0.19 to fetch and execute the Tiangong AI CLI tool. This is a vendor-owned package hosted on the official registry and is required for the skill's primary function.
  • [COMMAND_EXECUTION]: The wrapper script scripts/esg_search.sh executes shell commands to interact with the CLI tool. It includes configuration options to override the CLI entry point via environment variables, which is a standard developer feature.
  • [CREDENTIALS_SAFE]: Authentication is handled through specific environment variables (TIANGONG_ESG_APIKEY, TIANGONG_AI_APIKEY). The skill supports loading these from .env files located in the input file's directory, adhering to standard secret management practices.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: Processes user-supplied search queries and external JSON payload files (request_file, input_file) in scripts/esg_search.sh.
  • Boundary markers: Uses jq to parse and structure inputs before passing them to the CLI.
  • Capability inventory: Executes shell commands and has the ability to write search results to an output file path provided by the user.
  • Sanitization: Implements strict schema validation using jq filters to ensure ESG metadata filters are string arrays and date ranges are numeric, reducing the risk of malformed data injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 07:14 AM
Security Audit — agent-trust-hub — tiangong-kb-esg-search