| name | incident-response |
| description | Triage active incidents from alerts or failures into explicit hypotheses, mitigations, evidence, and next actions. Use when CTOX must react to outages, latency spikes, bad deploys, disk pressure, failing jobs, or repeated alert clusters and keep a crisp incident timeline grounded in concrete system state. |
| cluster | host_ops |
Incident Response
CTOX Runtime Contract
- Task spawning is allowed only for real bounded work steps that add mission progress, external waiting, recovery, or explicit decomposition. Do not spawn work merely because review feedback exists.
- The Review Gate is a quality checkpoint, not a control loop. After review feedback, continue the same main work item whenever possible and incorporate the feedback there.
- Do not create review-driven internal work cascades. If more work is needed, reuse or requeue the existing parent work item; create a new task only when it is a distinct bounded work step with a stable parent pointer.
- Every durable follow-up, queue item, plan emission, or internal work item must have a clear parent/anchor: message key, work id, thread key, ticket/case id, or plan step. Missing ancestry is a harness bug, not acceptable ambiguity.
- Rewording-only feedback means revise wording on the same artifact. Substantive feedback means add new evidence or implementation progress. Stale feedback means refresh or consolidate current runtime state before drafting again.
- Before adding follow-up work, check for existing matching internal work, queue, plan, or ticket state and consolidate rather than duplicating.
Only CTOX runtime store, ticket state, communication records, verification state, and direct live evidence count as durable incident knowledge. Ad hoc notes or markdown files do not count as durable knowledge by themselves.
Use this skill when there is a live user-visible or operator-visible failure that needs explicit stabilization work.
Do not use it for broad scope discovery or routine health review:
- use
discovery_graph when scope is unclear
- use
reliability_ops for health analysis without urgent containment
- use
change_lifecycle when the task is a planned rollback or controlled change
This skill uses the shared CTOX knowledge store via skill_key=incident_response.
Operating Model
Use CTOX CLI/API commands as the execution boundary. Do not execute embedded scripts/ helpers from this system skill; if incident capture or persistence lacks a CTOX command, add that command before relying on it.
Tool Contracts
incident.capture_raw
incident.store_capture
incident.store_graph
incident.query
incident.bootstrap_case
Harness Mining for Hypothesis Generation
When the incident touches harness behaviour (queue stalls, communication
failures, repeated state-machine rejections), use these CLIs to derive
hypotheses from the trigger ledger and preventive proofs rather than guessing:
ctox harness-mining causal --violation-code "<code>" --lookback 5 --limit 10
ctox harness-mining alignment --entity-type "<type>" --limit 10
What to read:
causal: by_violation_code[].predecessor_activity_lift[]. The activity
with the highest lift (and support ≥ 3) is your strongest causal
hypothesis — record it as a hypothesis with the lift value as its evidence
weight, not as a confirmed cause.
alignment: alignments[].moves[]. Every kind: "model" move marks a step
the spec demanded but the trace skipped — that is a concrete reparation
hypothesis. Use from_state and to_state from those moves to phrase the
mitigation, e.g. "force <entity> through <missing_state> before
<observed_state>".
A causal lift > 5.0 with support ≥ 5 is strong enough to anchor the
mitigation. Below that, treat it as a working theory only.
Workflow
- State the symptom, affected scope, and urgency.
- Capture first evidence before acting.
- Build a small hypothesis set.
For state-machine-shaped incidents, run
harness-mining causal and
harness-mining alignment first — they convert raw violation counts into
ranked predecessor hypotheses and concrete reparation moves.
- Prefer the smallest mitigation that matches the evidence.
If alignment proposes a missing intermediate state, prefer mutating the
spec or the producer — do not silently force the entity through.
- Persist an
incident_case, hypothesis_set, mitigation_action, and status_update.
Include the top three predecessor activities from causal (with their
lift) inside hypothesis_set evidence.
- If a real mutation is needed, hand the next work step to
change_lifecycle.
Operator Feedback Contract
Answer for the operator first, not for persistence details.
Use these exact headings:
**Status**
**State**
**Scope**
**Autonomous Actions**
**Escalation**
**Current Findings**
**Next Step**
State must be one of:
proposed
prepared
executed
blocked
If you stabilized part of the incident but did not finish the job, do not imply silent continuation. Say the incident is still open and point to the durable next-work record.
Completion Gate
Do not finish the reply until all of the following are true:
- all seven headings are present
State is explicit
- mitigation and evidence are clearly separated
- if the incident remains open, a durable next work step exists in queue or plan state
- if no durable next work step exists yet, the reply stays
blocked
Guardrails
- Do not claim root cause from one symptom.
- Keep mitigations narrow.
- Record evidence and action sequence.
- If the incident is not resolved, leave an explicit next work step.
Resources