Sentinel Workflow Manual

Appearance

Connected journeys ​

Manual home · Role guides · Coverage · Gaps

Manual conventions and recovery apply to every workflow. Evidence level: code inspected; related tests were inspected, not executed.

flowchart LR
  A[Operator configures tenant and source] --> B[Ingest redacted interaction]
  B --> C[Reviewer evaluates and submits]
  C --> D[Manager resolves disputes]
  D --> E[Coach activates plan]
  E --> F[Agent completes assigned learning]
  F --> G[Manager checks measured outcomes]
  G --> H[Deliberate knowledge and scoring revision]

XFN-01 — Trace an interaction from ingestion to measured improvement ​

Status: Implemented. Verification: code inspected; no runtime completion claim.

Purpose: trace an interaction from ingestion to measured improvement.

Actors and access: Integration operator → reviewer → manager/coach → agent, each with the domain permissions.

Prerequisites: Configured ingestion, published scorecard and eligible tenant users.

Entry point: Start with Interactions/integration setup; follow linked domain workflows.

  1. Ingest via INT-02/03/04/05; verify redacted evidence and analysis before allocating QMS-01.
  2. Review and submit via QMS-02/03/04. Resolve disputed outcomes through corrections, re-review or calibration/adjudication rather than editing submitted records.
  3. Create/activate coaching with LRN-05/06 and assign training with LRN-03; compare sufficient outcomes or capture an improvement baseline/pilot with JNY-05/06.

Result: Evidence and accountable human decisions connect source interaction to measured follow-up.

Exceptions and recovery: Each handoff is conditional, not an automatic chain. Missing canonical correlation, sparse metrics, worker failures and current agent UI gaps can interrupt the journey; retain linked IDs and inspect the relevant gap/recovery instructions.

Source evidence: services/api/src/app.ts; services/api/src/audits/routes.ts; services/api/src/coaching-plans/routes.ts; services/api/src/workspace/routes.ts.

XFN-02 — Prepare a tenant for controlled operational use ​

Status: Implemented. Verification: code inspected; no runtime completion claim.

Purpose: prepare a tenant for controlled operational use.

Actors and access: Platform operator, tenant owner and manager.

Prerequisites: An agreed tenant/authentication provider and deployment dependencies.

Entry point: Platform → workspace setup → scoring/integrations.

  1. Provision tenant/access with OPS-07 and ACC-02/05/07/08; configure timezone/retention/transcription with ACC-09.
  2. Connect sources/models through OPS-01/02/05; author and publish evaluation definitions through EVA-01–07.
  3. Validate a non-production ingestion/review example and inspect privacy, roles, notifications and outcomes before operator-approved rollout.

Result: A repeatable readiness checklist connects configuration to a demonstrable vertical workflow.

Exceptions and recovery: Creating configuration is not production acceptance. Journey coaching remains gated; external deliveries/provider calls require their real operator context. This manual audit performs no onboarding writes or live test delivery.

Source evidence: README.md; services/api/src/app.ts; docs/journey-intelligence-rollout.md.

XFN-03 — Connect published knowledge to scoring and learning ​

Status: Implemented. Verification: code inspected; no runtime completion claim.

Purpose: connect published knowledge to scoring and learning.

Actors and access: Knowledge author/publisher → scoring administrator → reviewer → learner.

Prerequisites: Published article revision and supported linked scoring definitions.

Entry point: Knowledge → Atoms → Molecules → Scorecards → reviews/training.

  1. Create/publish knowledge through KNO-02/03; create linked Atoms and publish their Molecule/scorecard composition.
  2. Review an interaction using its pinned knowledge selection and evidence through QMS-02/04.
  3. Use adherence/report evidence and feedback to revise knowledge/scoring deliberately; announce changes and assign learning/training where needed.

Result: Published knowledge has traceable scoring and learning use without rewriting historical citations.

Exceptions and recovery: Publishing newer knowledge does not refresh existing pinned definitions. New Atom/Molecule/scorecard versions and assignments may be needed. Announcement reads, quiz completion and observed quality improvement are different evidence types.

Source evidence: services/api/src/knowledge/routes.ts; services/api/src/audits/knowledge-selection.ts; services/api/src/reports/routes.ts.