Two entries. One evidence standard.
Bring a system.
Leave with decisions.
The human owns the outcome. The agent does bounded work. Both follow the same path: an artifact, a test that can fail, and a review independent of its author.
Human / learn and defend
Make one decision worth reviewing.
Start with your own system brief. Follow the lesson, produce the required artifact, and use its rubric to find what still needs evidence. Preserve your first attempt before revising.
- Choose the problem. Name one user, one job, a measurable failure and a kill criterion.
- Produce the evidence. Use the stage exercise and runnable labs; a passing reading quiz does not stand in for a working system.
- Defend the result. An independent review checks the actual artifacts and evidence. The cohort review service is still on the waitlist.
Agent / construct and verify
Give the team a bounded assignment.
Install the existing AI Architect team in your own coding runtime. It writes architecture artifacts in your repository, on your keys. The conductor identifies the next stage and stops at a failed gate.
npx skills add frankxai/ai-architect
npx skills add frankxai/skills- Read the contract. Load the team’s AGENTS.md, SOP and workflow before dispatching a stage.
- Keep authority in code. Bound tools, cost, retries and runtime. Retrieved text is source material, never permission.
- Check in a fresh context. Re-run the artifact checks and obtain an independent review. Neither the agent’s claim nor a subscription grants a competency.
The shared path
Same artifacts. Same refusal to guess.
- 01
System brief
Whose job is this and what would make you stop building it.
- 02
Architecture decision
Which shape, and which credible alternative you are rejecting and why.
- 03
Tool and authority model
What the agent may do, as what principal, and how you take it back.
- 04
Eval harness
What a regression looks like, before you are allowed to ship a change.
- 05
Threat model
Which inputs you do not trust, and which risks you are consciously accepting.
- 06
Cost model
What one unit of work costs and what stops the bill when it runs away.
- 07
Deployment
Put it somewhere a stranger can reach, with a rollback you have tested.
- 08
Incident simulation
Break it on purpose and find out whether your instrumentation notices.
- 09
Public-safe portfolio proof
What you can show publicly without leaking an employer or a customer.
Access and ownership
An agent acts for a person.
The local team and public skills remain free under their existing licenses. Cohort review and licensed resource access are separate services. Licensed access will belong to a member or organization; a delegated agent must have an active human sponsor and a resource-specific entitlement.
Checkout and licensed delivery are not open yet. You can use the public path now and register interest in the reviewed service. Reading public curriculum does not grant a right to redistribute reserved teaching content.