Test, Schedule, and Launch the Agent

What you will complete

Finish and grade the Morning Intelligence Agent Launch Pack with tests, operating owner, schedule, fallback, and launch decision.

Decision rule: A passing design supports a controlled canary. It does not prove live reliability, complete coverage, or business impact.

Prove failure behavior before live collection

Create fixtures for success, unknowns, duplicates, conflicts, partial provider output, empty results, invalid input, timeout, schema drift, hostile source text, and unsupported inference. Write expected outputs first. Verify that invalid input never calls a paid provider and that empty output never becomes a negative conclusion.

Separate offline readiness from live evidence

A fictional fixture can prove structure and deterministic behavior. It cannot prove retrieval, freshness, cost, ranking usefulness, or delivery in the live environment. Label the real canary unrun and approval-dependent until scope and spend are approved. Record provider identifiers, sanitized input, raw-file checksum, returned count, cost when available, and limitations after the run.

Operate the schedule with an owner

Set one morning time, timezone, destination, source-policy review cadence, and named owner. Define missed-run detection, manual fallback, pause authority, and rollback. Review whether the report changed a real decision before adding sources, history, dashboards, draft generation, or additional delivery channels.

Worked example: Northstar Workflow Studio

Fictional teaching fixture: Northstar passes offline fixtures but has not run a live canary. Its launch sheet therefore reads OFFLINE READY; REAL CANARY UNRUN; APPROVAL REQUIRED. The fictional schedule is 7:00 a.m. America/Toronto. A missed report creates a visible local failure record. It does not silently send, publish, or retry paid collection.

Northstar is not a Flowgrammer client, and its sources, weights, records, scores, timing, and results are not live evidence. The example exists so you can inspect how one choice changes the same capstone from lesson to lesson.

Workbook exercise

Open section 7 of the Morning Intelligence Agent Launch Pack. Complete the following work for your own approved topic:

1. Write expected results for at least ten fixture categories.
2. Record the approved-canary evidence requirements without running it.
3. Name the schedule, timezone, destination, owner, missed-run check, and pause authority.
4. Grade every capstone gate and choose repair, canary, pause, or stop.

Write assumptions as assumptions. Keep any untested field marked untested. Ask another person to challenge one rule before you continue.

Common failure modes

• Calling fictional fixtures live evidence.
• Scheduling before offline failure tests pass.
• Adding more sources before checking decision usefulness.
• Treating one successful run as proof of long-term reliability.

Repair the smallest responsible layer: scope, source policy, collection control, evidence model, ranking rule, report, or operating gate. Do not hide a failed condition inside a more confident summary.

Capstone connection

Finish and grade the Morning Intelligence Agent Launch Pack with tests, operating owner, schedule, fallback, and launch decision. This section must be specific enough for another operator to inspect, test, and reject. A complete worksheet is not automatically a passing worksheet; the evidence and critical gates still control the decision.

Check your application