the-cohort-tracker
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data provided by users (such as CRM exports or pasted tables), which is a common ingestion point for indirect injection attacks. However, the skill includes robust mitigations.
- Ingestion points: Raw data provided via user input (Step 3 in SKILL.md).
- Boundary markers: The
references/agent-security.mdfile contains specific 'Five Rules' instructing the agent to treat all retrieved content as data, never instructions, and to explicitly quote and report any detected injection attempts. - Capability inventory: The skill writes analysis snapshots to the local
.agents/gtm-run-state.mdfile system path. - Sanitization:
references/agent-security.mdRule 3 mandates HTML-escaping for all interpolated values in emitted markup to prevent XSS. - [DATA_EXPOSURE]: The skill manages persistent state through snapshots written to
.agents/gtm-run-state.md. It includes proactive security measures to prevent sensitive data exposure. - The instructions explicitly forbid the persistence of 'special-category data' or 'raw personal data'.
references/agent-security.mdRule 2 provides strict protocols for redacting hardcoded secrets or credentials found in inputs.
Audit Metadata