extruct-api
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting data from external web research and third-party platforms (LinkedIn) and interpolating it into subsequent AI prompts.\n
- Ingestion points: Data fetched from the Extruct API via the CLI, such as company profiles and LinkedIn activity, as described in
SKILL.mdandreferences/column-guide.md.\n - Boundary markers: The prompt templates (e.g., for
pricing_modelorprioritizationinreferences/column-guide.md) use curly-brace interpolation for data but do not include explicit security delimiters or instructions to ignore embedded commands.\n - Capability inventory: The bundled CLI (
<extruct_api_cli>) provides capabilities for network communication and managing data structures (tables, rows).\n - Sanitization: There is no evidence of data sanitization or filtering of external content before it is re-processed by the agent.\n- [COMMAND_EXECUTION]: The skill requires the execution of a local bundled script to perform its primary functions.\n
- Evidence:
SKILL.mdprovides detailed instructions on resolving the path toscripts/extruct-apiand executing it with various arguments likeauth,tables, anddeep-search.
Audit Metadata