Clear Promise
After finishing, you can design one narrow reviewed helper with a charter, context, permission map, test set, receipt, and human gate.
$49 standalone Build
Basic Agents
Design your first useful AI helper with a narrow role, trusted context, safe permissions, test cases, and a human review gate. The goal is not autonomy theater. The goal is a helper boring enough to trust.
Buyer Clarity
Use this quick check to decide whether this Build is worth buying by itself.
Who This Is For
Learners who want agents but need to know the difference between a reusable assistant pattern and an autonomous system that can act without review.
First 15 Minutes
Pick one recurring task, decide whether it is agent-shaped, and write what the helper may read, draft, summarize, or suggest.
Quality Checks
The helper has a narrow job, bounded inputs, explicit non-actions, test cases, failure signals, review gate, and a receipt of what changed.
How access turns into a first artifact.
This is a standalone Build. Checkout creates a clean learning handoff: what you bought, where to start, and which first artifact to build before opening the full Build pack.
1. Start Checkout
Use checkout when this Build matches the current work bottleneck.
2. Receive Access Status
The receipt confirms the Build, price, access link, start file, refund window, and support boundary.
3. Open The Start Path
After checkout, start with the Build room or paid-buyer start file before opening the Build zip.
4. Build The First Artifact
Use the first action, worked example, template, and rubric to produce a useful first draft.
Why This Stands Alone
This Build is worth buying by itself when a recurring task might deserve a helper, but the learner needs a safe boundary before building anything. The payoff is one narrow, reviewed agent spec rather than a vague autonomous assistant.
Immediate Payoff
Classify one recurring task as prompt, template, basic agent, workflow, automation, or human-only hold before overbuilding.
Finished Artifact
Leave with a First Agent Pack: charter, context packet, permission map, test set, receipt template, quality score, and launch-mode decision.
Quality Gate
Use the agent quality rubric and five-case test to catch missing context, overreach, unsafe permissions, weak receipts, and unclear human gates.
The Build stands alone because the buyer can design one useful reviewed helper without buying the full Agentic Workflows class or setting up automation.
Buy This If
This Build is for the first serious agent conversation: when a recurring task might deserve a helper, but you want the role, context, permissions, and tests designed before the tool gets power.
Good Fit
- You have one recurring task that could be drafted, reviewed, sorted, flagged, or reported by an assistant.
- You want a practical agent spec you can use in ChatGPT, Claude, Gemini, Codex, Claude Code, Antigravity, or another surface.
- You need a permission map before connecting tools, files, inboxes, calendars, or MCP access.
- You want tests and receipts before trusting the helper.
Not The Right First Step
- You want a broad autonomous worker with open-ended permissions.
- You have not identified a recurring task, output, or review gate.
- You want the agent to send, delete, buy, publish, schedule, or change records without approval.
- You need to map a whole multi-step process first; that belongs in Agentic Workflows.
Your First 30 Minutes
A learner should get a practical win before the Build starts to feel theoretical. Start with one bounded artifact, use low-risk material, and stop before widening the system.
First Artifact
An agent-or-not triage card for one recurring task.
- Choose one recurring task that could benefit from a standing helper: status drafts, meeting briefs, report review, research triage, or follow-up drafts.
- Classify it as prompt, template, basic agent, workflow, automation, or human-only hold.
- Write what the helper may observe, decide, do, report, and where it must stop for human approval.
Early Win
You know whether the task deserves an agent shape before giving any tool real authority.
Open first: FIRST_ARTIFACT_STARTER.md, then AGENT_NOT_AGENT_DECISION_CARD.md and PERMISSION_MAP.md.
The Lesson Sequence
The sequence lowers the temperature around agents. The durable skill is role design: what the helper may observe, decide, do, check, report, and stop.
| Lesson | Outcome | Class Asset |
|---|---|---|
| 1. Is this really an agent? | Classify a recurring use case as prompt, template, basic agent, workflow, automation, or human-only hold. | Agent/not-agent lesson |
| 2. Choose the first candidate | Select one narrow, reviewable helper role or safely reject the agent idea. | Decision card |
| 3. Map the loop | Define how the agent observes, decides, acts, checks, reports, and stops. | Agent loop map |
| 4. Write the charter | Name the role, job, inputs, outputs, non-goals, and human review gate. | Charter template |
| 5. Attach context and permissions | Set source authority, freshness rules, exclusions, allowed actions, forbidden actions, and approval gates. | Context packet |
| 6. Test before launch | Run normal, messy, missing-context, boundary, freshness, and format cases before choosing launch mode. | Agent test set |
| 7. Package the first reviewed agent | Assemble the final spec, launch prompt, receipt template, rubric score, and launch decision. | First agent spec |
Flagship Example
The flagship example builds a draft-only project status agent. It is useful precisely because it does not send updates, edit sources, change priorities, or invent progress.
Weak Agent Request
"Make me an agent that handles my project updates."
- The job is too broad.
- Inputs and source authority are unclear.
- Sending, editing, and priority decisions are not gated.
Strong First Agent
"Draft a Friday project status update from the notes I provide, flag blockers and unclear ownership, and return a receipt. Do not send, publish, edit files, change priorities, or mark work complete."
- The role is narrow and repeatable.
- The output is reviewable.
- The permission boundary is explicit.
Open the full walkthrough: Full Example: Project Status Agent. For a first-hour proof path, use Sample Lesson: Is This Really An Agent?, then compare it with First Agent Spec Preview and Agent Test Set Preview.
Included Downloads
The learner leaves with a practical agent pack that can be reviewed before any tool gets real authority.
Course Room
- Private paid learner room after access
- Start file, first 15-minute action, lesson path, download shelf, artifact checklist, and completion gate.
Core Class Files
- Start here file
- Student path
- First artifact starter
- First agent capstone
Agent Design Files
- Agent pattern card
- Permission map
- Agent receipt template
- Completed first agent example
Review And Delivery
- Agent quality rubric
- Practice labs and common mistakes
- Workshop adaptation
- Download Build pack
Live delivery use: this Build can become a first-agent clinic, permission-mapping lab, or reviewed-helper workshop. The live version should leave each participant with one narrow agent spec, test set, and human gate instead of a promise of autonomous work.
Capstone And Rubric Standard
A learner completes the Build when the First Agent Pack is narrow, tested, gated, portable, and assigned an honest launch mode.
The Finished Packet Includes
- Agent/not-agent decision and reason.
- Agent charter with non-goals.
- Context packet with source authority and freshness rules.
- Permission map with allowed, forbidden, dry-run, and approval actions.
- Five-case test set and launch receipt.
Common Mistakes To Catch
- Calling every reusable prompt an agent.
- Starting with connected tools before role design.
- Skipping messy, stale, missing-context, or forbidden-action tests.
- Letting memory or records change silently.
- Launching without a receipt a human can review.
Next class: Agentic Workflows, if the tested agent is one piece of a larger repeatable process. If the agent is not yet stable, repeat the test set before widening its role.
Access, Support, And Refund Boundary
This is a standalone Build, not a promise of unlimited consulting or external certification. Learners should know exactly what access includes before payment is accepted.
Access
Self-serve access should include at least 12 months of access to the Build materials and minor updates during that access period.
Support
Access help, missing-file fixes, broken links, and basic start-path confusion are included. Custom implementation, done-for-you work, live consulting, and artifact review are separate offers.
Refund Window
Confirm the 7-day fit-based refund window before payment is accepted. The window should cover duplicate charges, access problems, or a genuine mismatch with the Build description.
Tool choice stays platform agnostic. Completion means you build the artifact and review it against the rubric; it is not external accreditation or a guarantee of business, career, or productivity outcomes. For file, agent, workflow, or workshop work, start with low-risk material and keep human review in the loop before external use. Read the Buyer FAQ for fit, support, refund, safety, and completion boundaries.
Return To Current Offers
This Build is part of the six-Build launch path. Enroll through checkout, open the course room, and build the First Agent Pack before widening the helper's role.