Starlight IntelligenceStarlightAcademy · Operator Lab

Starlight Workshop

Bring one workflow. Build the operating packet.

A focused workshop for turning recurring AI work into explicit roles, evidence, and human decisions.

A concept for the existing Operator Lab method—not a certification, autonomous-agent promise, or confirmed cohort offer.

Inspect the packetShare a workflow
  • One recurring workflow
  • No site storage
  • Human-owned decisions
Workshop packet 01Educational example

Workflow review

Invoice exception review

Ready to test
AI role
Evidence collector — never approves payments
Evidence gate
Source record · exception reason · owner · timestamp
Human decision
Finance owner approves every external message and payment decision
Run receipt
Objective · source trail · checks · status · next action
sample_operator_lab_001Inspect source

One governed loop

  1. 01Bound
  2. 02Map
  3. 03Test
  4. 04Hand off

Inside the workshop

Four moves from prompt to operating practice.

The work stays narrow enough to test. Every step creates an artifact an operator can inspect, revise, and keep.

  1. 01

    Bound the work

    Name one recurring job, one owner, and the conditions that stop the system.

    Operating contract
  2. 02

    Map the evidence

    Decide which sources matter, what expires, and which uncertainty must stay visible.

    Evidence map
  3. 03

    Test the route

    Give AI the smallest useful role, then pressure-test exceptions and human review points.

    Evaluation rubric
  4. 04

    Hand off the decision

    Leave the owner with a receipt that explains what happened and what decision is needed.

    Run receipt

The boundary is part of the product

AI prepares. Evidence accumulates. A person decides.

Bring

  • One recurring, non-confidential workflow summary
  • A named owner and a real review moment
  • Redacted examples or low-risk fixtures

Keep human-owned

  • External sends, publishing, spend, and commitments
  • Exceptions involving sensitive or uncertain data
  • The final decision and accountable handoff

Portable by design

The packet survives the next tool change.

The method produces plain, inspectable files. No private memory or production system is exposed by this public learning surface.

AGENTS.md

Role, authority, boundaries, and stop conditions

MEMORY_MAP.md

Context that persists, expires, or stays private

SKILL.md

A portable procedure for the bounded workflow

EVAL_RUBRIC.md

Evidence required before the work moves forward

RUN_RECEIPT.json

A decision-ready record of the run

Workshop concept

Start with the workflow worth governing.

Share only a non-confidential summary. Fit, delivery, dates, capacity, and any commercial terms are confirmed directly before commitment.

Share a workflow for considerationRun the browser-local triage insteadEmail opens in your client · the site stores nothing