client-health-check
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data to perform its primary function.
- Ingestion points:
SKILL.mdinstructs the agent to pull facts from 'email threads, invoices, meeting notes' to inform the health scoring process. - Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the analyzed materials, which could allow maliciously crafted emails to influence the scoring logic.
- Capability inventory: The skill does not define or use any external scripts, subprocesses, or network tools, limiting the potential impact of an injection.
- Sanitization: No sanitization or data validation steps are implemented for the ingested content.
- [NO_CODE]: The skill consists entirely of natural language instructions and does not include any executable scripts, binaries, or configuration files that could perform actions outside of the agent's core reasoning capabilities.
Audit Metadata