find-linkedin-profile
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
claycommand-line interface (CLI) to perform person searches and execute data routines (e.g.,clay search filters-mode run,clay routines runs start). These are legitimate vendor-specific operations designed to interface with the Clay platform. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from search results and enrichment payloads. It mitigates potential risks by requiring strict validation signals (Name, Employer, and Experience History) before reporting data to the user.
- Ingestion points: Clay search result records (SKILL.md, Step 2), 'Enrich Person' function payloads (SKILL.md, Step 3), and local answer sheets (SKILL.md, Declared inputs).
- Boundary markers: The instructions mandate a 'find → validate → recover' ladder and explicitly forbid fabricating URLs or reporting unvalidated hits.
- Capability inventory: The skill uses subprocess calls to the
clayCLI and initiates network operations viaclay routines. - Sanitization: Implements a three-signal validation score (Step 3) to ensure the ingested data matches the expected identity before it is accepted.
Audit Metadata