crisis-holding
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates solely as a set of instructions for text generation based on user-provided incident facts. It lacks access to tools, network capabilities, or the ability to modify the host system, significantly limiting any potential for exploitation.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests untrusted data in the form of incident summaries and facts, it includes robust defensive instructions such as a hard legal-counsel gate and a requirement to draft only from confirmed facts. Because the skill has no actionable capabilities (such as network access or file-writing), the risk of indirect prompt injection is minimal. Evidence chain:
- Ingestion points: Intake fields in SKILL.md (Incident summary, known facts).
- Boundary markers: The structured intake table format and the 'Anti-slop principle' which enforces strict adherence to facts.
- Capability inventory: None (no tool usage, subprocesses, or network access).
- Sanitization: Instructions to omit unverified information and the 'Hard legal-counsel gate' that prevents output when triggers are met.
Audit Metadata