zmm-concentration
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, specifically financial records, customer lists, and revenue details provided by the user. While this is the intended functionality, the instructions do not specify boundary markers or sanitization steps to prevent malicious instructions hidden within that data from influencing the agent's behavior.
- Ingestion Points: The skill requests a 'revenue list sorted by customer' and 'customer source' data in Phase 0.
- Boundary Markers: There are no specific delimiters or instructions to ignore embedded commands within the ingested financial data.
- Capability Inventory: The skill is instructed to read local configuration files (
config.yaml) and read/write to memory paths ({config.paths.memory}/zmm-concentration/). - Sanitization: No explicit escaping or validation logic is defined for the external data provided by the user.
Audit Metadata