tam-building

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from LinkedIn Sales Navigator, which is an external and potentially untrusted source.
  • Ingestion points: Data enters the agent's context through the salesnav_companies model in models/salesnav-companies.ts using the fetchAccountSearch action.
  • Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded commands within the fetched LinkedIn company data.
  • Capability inventory: The skill has the capability to perform database operations (model.upsert) and call external business matching/enrichment APIs via the cargoDb connector.
  • Sanitization: The promoteCompany workflow in plays/promote-to-accounts.ts implements basic validation using Zod for input structure and checks for the presence of a website before promotion.
  • [EXTERNAL_DOWNLOADS]: The skill requires the installation of external vendor packages and resources.
  • Evidence: The SKILL.md file specifies the installation of the @cargo-ai/cli package and references fetching the deploy-cookbook from the official getcargohq GitHub repository.
  • [COMMAND_EXECUTION]: The documentation includes multiple examples of shell commands for the user to execute using the cargo-ai CLI tool.
  • Evidence: Instructions in README.md and SKILL.md guide the user to run cargo-ai orchestration action execute to count searches and npx skills add to scaffold the environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 07:44 AM
Security Audit — agent-trust-hub — tam-building