triage
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, such as monitoring alerts and user reports, which could potentially contain malicious instructions intended to influence the agent's coordination of incident response.
- Ingestion points: Untrusted data enters the agent context through Nexus (incident routing), monitoring alerts, and user reports, as defined in the SKILL.md and reference/response-workflow.md files.
- Boundary markers: The skill relies on platform-level agent role boundaries (_common/BOUNDARIES.md) and includes specific runbooks (e.g., A3 in reference/response-workflow.md) to detect and deny prompt injections reaching external write operations.
- Capability inventory: The skill coordinates sensitive actions by delegating to other agents, including Builder (implementing fixes), Gear (executing rollbacks), and Sentinel (security analysis).
- Sanitization: The skill employs structured stakeholder communication templates and mandates legal-review hooks for external communications involving sensitive data, as detailed in reference/incident-communications.md.
Audit Metadata