Verifiedon 2026.7.1-2
View this course’s maintenance history
Practical outcome
What this gives you
- Decompose a production workflow into accountable roles, artifacts, and dependency gates
- Set budgets, deadlines, stop conditions, and human escalation before work starts
- Operate durable jobs with idempotent inputs, run history, and a recovery decision
- Evaluate output against written acceptance criteria and publish only after review
Preflight
Before you start
- CLI Essentials and Operator Diagnostics
- Security Hardening and Threat Boundaries
Syllabus
Work through the map
- Decompose work into outcomes and dependency gates20 min
- Assign role boundaries and inspectable handoffs20 min
- Set goals, budgets, stop conditions, and escalation20 min
- Operate durable jobs with retry and idempotency gates20 min
- Evaluate output and enforce review gates20 min
- Capstone: run a human-gated release-intelligence workflow20 min
- Final assessmentAfter coursework
Course details and supporting evidence
What you will operate
A release-intelligence workflow that turns a bounded source list into a draft bulletin. A researcher collects primary-source URLs, a verifier checks each claim against a receipt, an editor writes a concise draft, and a human publication owner decides whether anything is released. Every handoff has an owner, a visible artifact, a deadline, a stop condition, and a recovery path.
This is not a recipe for unattended work. More agents only help when the work can be divided without hiding authority, cost, or failure.
Course map
- Turn a vague request into accountable outcomes and a dependency graph.
- Establish role boundaries, isolated workspaces, and safe handoffs.
- Set goals, budgets, stop conditions, and escalation rules.
- Run durable work with idempotency, retry classification, and operational signals.
- Assess evidence and require review gates before any consequential action.
- Run the capstone, inspect the audit trail, and conduct a blameless improvement review.
Learner-produced visual evidence
- Dependency graph: draw source intake → research → claim review → draft → human publication gate. Use solid arrows for required dependencies and a red stop node for missing evidence, budget breach, or unapproved publication.
- Operations dashboard: make a redacted mock dashboard showing queue depth, oldest in-flight task, retry count by failure class, budget consumed, review backlog, and last successful run. Show thresholds and the named escalation owner beside each signal.
- Handoff card: include task ID, role, input fingerprint, acceptance criteria, owner, deadline, cost ceiling, status, retry count, and rollback/stop action.
Source receipts and version boundary
Technical claims in these lessons are tied to OpenClaw commit ec53d0eaae19b0e54522bc91ab06234427acafd9 and npm 2026.7.1-2. Recheck the linked primary documentation before changing an agent boundary, scheduler, command, or retry policy.
Launch video decision — no videos
This launch intentionally includes no third-party video. videoIds is empty, no player or VideoObject is rendered, and course completion does not depend on video. The Brian Casel candidate (bzWI3Dil9Ig) remains a dormant optional-curation record only: it cannot be embedded unless a future independent review records segment viewing, a current-docs cross-check, availability/embed evidence, privacy/rights review, and accessibility review. No transcript is reproduced.
Completion evidence
Keep an operational acceptance pack: dependency graph, role and authority matrix, sample durable-job definition, failure/retry tabletop, dashboard thresholds, two reviewed claims with source receipts, approval record, and postmortem note. Redact private URLs, identifiers, and credentials.