account-tier-scoring
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes vendor-provided CLI tools (
clay) to verify the workspace status and interact with enrichment routines. These operations are limited to environment verification and metadata retrieval. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted account descriptions. 1. Ingestion point: The 'book' input described in Step 1. 2. Boundary markers: Absent for AI judgment tasks in Step 2. 3. Capability inventory: Operations are limited to reading account data and calling managed vendor routines (
clay routines). 4. Sanitization: Absent for text descriptions, though numeric coercion is strictly applied for scoring weights. The risk is significantly minimized by the skill's requirement to decompose AI judgment into specific sub-scores and evidence phrases rather than allowing unstructured output.
Audit Metadata