Morning Intelligence Agent Launch Pack
A printable workbook for defining sources, collection controls, evidence records, ranking rules, report structure, tests, schedule, and launch gates.
Use this workbook with Build a Morning Intelligence and Trend-Watching Agent. Complete one section after each lesson.
Included
- Agent charter and v0 boundary
- Source policy and watchlist
- Collection configuration and cost controls
- Evidence, date, unknown, and dedupe contract
- Trend-ranking rubric
- Morning report and delivery contract
- Fixture suite, schedule, operations, and capstone gates
How to use it
Choose one named reader and decision before adding sources. Complete the matching section after each lesson, retain the raw source beside every normalized record, and keep collection failures visible in the report. Test duplicates, stale items, missing dates, blocked sources, injected instructions, and empty-result runs before scheduling. Review the finished launch pack with the person who owns the briefing decision and the person responsible for operating failures.
Before you schedule the agent
Confirm that every source is approved, retrieval limits and expected costs are recorded, and failures reach a named operator. Check that the report distinguishes observations, interpretations, recommendations, and unknowns. Run every applicable fixture after the last repair. Keep delivery internal until a reviewer approves the destination, audience, cadence, pause rule, and recovery process.
All Northstar material is fictional. A live canary remains unrun and approval-dependent. The workbook does not authorize paid collection, private-data access, sending, publishing, outreach, or production deployment.
Choose a decision before choosing more sources
Use this nine-page PDF to define the brief’s reader and the decision it supports. A founder scanning competitor changes needs a different brief from a service team tracking operational alerts. Choose a small starting source list and record why each source earns a place.
For a practice run, take several public items about one topic. Record the original publication date and link, separate a new announcement from commentary about it, and explain why the item matters to the intended reader. Several articles repeating one announcement do not establish a trend.
Review a sample before scheduling delivery
Ask the reader to check one sample brief for relevance, repetition and unsupported interpretation. Keep the source visible beside each conclusion. If nothing material changed, a short “no qualifying updates” result may be more useful than filling space with marginal items.
Decide who can change the source list, where uncertain items go and how a broken source is reported. Schedule the brief only after those choices are clear. Use the morning intelligence course to work through the underlying workflow; the PDF is a planning companion rather than a deployed monitoring service.