Start a diagnosis
Diagnosis starts from an authorized incident, not from an empty run form. This keeps the project, workspace, incident, provider, and evidence relationship visible before work is accepted.
- Typical time
- 5–10 minutes plus queue time
- You need
- An incident with ready, unexpired evidence
- Outcome
- An accepted advisory diagnosis request
Step 1: Open an eligible incident
Section titled “Step 1: Open an eligible incident”In Incidents, select the intended project and workspace, open View details, and inspect Evidence. Start diagnosis is disabled when there is no ready, unexpired evidence manifest.
Diagnosis start 1 of 4 · Screenshot placeholder
Confirm the incident is eligible
Show a synthetic incident drawer with ready evidence and Start diagnosis available.
Step 2: Choose Start diagnosis
Section titled “Step 2: Choose Start diagnosis”Choose Start diagnosis. ScrinOps opens a guided start drawer and carries the server-readable incident context into Runs.
The Runs page deliberately shows Start from Incidents as disabled. This prevents a user from inventing an incident ID or starting outside the evidence review flow.
Step 3: Review Authorised scope
Section titled “Step 3: Review Authorised scope”Confirm the displayed Project, Workspace, Incident, and provider Source. If any value is unexpected, choose Cancel and return to Incidents.
Diagnosis start 2 of 4 · Screenshot placeholder
Review the authorized diagnosis scope
Step 4: Check Preflight readiness
Section titled “Step 4: Check Preflight readiness”Confirm:
- tenant and workspace scope is ready;
- at least one eligible evidence manifest is included;
- raw provider data and credentials are excluded; and
- any unavailable evidence manifests are visibly excluded.
The backend remains authoritative and can reject a request even when the browser preflight looks ready.
Diagnosis start 3 of 4 · Screenshot placeholder
Check diagnosis preflight
Show synthetic ready scope, eligible evidence count, excluded unavailable evidence, and redaction boundary.
Step 5: Add optional repository context
Section titled “Step 5: Add optional repository context”Under Optional repository context, choose an existing Repository binding (optional) or leave No repository context selected. No-Git diagnosis is supported by the interface.
A repository binding adds only server-authorized context. The browser cannot select paths, tools, commands, prompts, credentials, model, runner profile, timeout, or placement.
Step 6: Start read-only diagnosis
Section titled “Step 6: Start read-only diagnosis”Choose Start read-only diagnosis once. The request uses an idempotency control to prevent duplicate acceptance. A success message means diagnosis was accepted or queued; it does not mean diagnosis completed.
Diagnosis start 4 of 4 · Screenshot placeholder
Start the read-only advisory run
Show Start read-only diagnosis with no repository context or a synthetic approved binding, followed by an accepted message.
Safety boundary
Section titled “Safety boundary”ScrinOps diagnosis is advisory and read-only. It cannot apply Terraform, merge code, change provider resources, alter incidents, or approve remediation. Untrusted logs and repository content do not expand the tool or scope boundary.
Check your result
Section titled “Check your result”The request appears in Runs with the expected incident and scope and has an authoritative queued, deferred, starting, running, or terminal status.
Common blockers
Section titled “Common blockers”- Start diagnosis disabled: no ready, unexpired evidence is available.
- Scope is wrong: cancel and reopen the incident in the correct project and workspace.
- Not authorized: the authenticated role, plan, or current capability does not permit the run.
- Runner unavailable: capacity or the deployed runtime contract is not ready.
- Repeated click: wait for the first request; do not create duplicates.
The source-defined drawer and API wiring do not prove a live runner accepted the request. Confirm that only from the runtime run record in the intended environment.