bs-first-customer-finder
Pass
Audited by Gen Agent Trust Hub on Jul 15, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: No evidence of instructions designed to bypass agent safety filters or override system constraints. The instructions focus on maintaining high standards of evidence and rejecting weak matches.
- [DATA_EXFILTRATION]: The skill explicitly forbids the collection of private data, such as emails or phone numbers, and prohibits bypassing access controls (Rule 6). Outreach actions are restricted to manual processes requiring explicit authorization (Rule 8).
- [COMMAND_EXECUTION]: No shell commands, script generation, or subprocess execution patterns were identified in the instructions or references.
- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to process untrusted external data from the web, it implements a robust 'HARD-GATE' for eligibility and a 'Adversarial Review Panel' (Step 5) to ensure all findings are grounded in evidence and to prevent the agent from obeying instructions that might be embedded in scraped content.
- [METADATA_POISONING]: All metadata fields, including the description and patterns section, are consistent with the stated purpose of the skill.
Audit Metadata