playbook-company-name-cleaning

Pass

Audited by Gen Agent Trust Hub on Aug 23, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by processing untrusted company name strings from lead databases.
  • Ingestion points: company_name_raw input in SKILL.md and implementation files.
  • Boundary markers: The LLM prompt uses clear labels to isolate input data.
  • Capability inventory: The system allows for JavaScript and Python script execution for data processing.
  • Sanitization: The skill implements a robust, deterministic invented-word guard (substring check) ensuring the AI output is strictly derived from the input, mitigating the risk of instructions embedded in company names being executed.
  • [COMMAND_EXECUTION]: The clay-workflow.md file contains CLI command examples for the Clay platform. These are standard developer tools and do not pose a security risk as they are used for legitimate routine management.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 23, 2026, 02:36 PM
Security Audit — agent-trust-hub — playbook-company-name-cleaning