Black Belt Part B — Push the Craft
Status: drafted · Time: 8 min · Audience: platform-builder Outcome: Understand what Part B teaches, what Part A muscle memory it assumes, and the order to walk it in.
Part B promise. By the end, you can keep skills small with progressive disclosure, design memory systems for long-running agents, run prompt evals on shipped skills, instrument cost and observability at team and org scale, and tune effort settings + model routing + fall-backs.
Prerequisite. Part A complete (Quest B-1 claimed). Part B reads thin without the platform-builder craft from B.1–B.6. Time budget. ~3h for the five modules; ~8h for Quest B-2. Plan over a week or two.
What you’ll learn
Section titled “What you’ll learn”Part A taught the platform layer: MCP servers, skill packs, Cowork marketplace entries, custom agents, multi-agent orchestration, tool design. Part B sharpens the craft on top of that platform: progressive disclosure for skills that stay small, memory systems for long-running agents, prompt evals that turn taste into measurement, cost-and-observability at scale, and the per-call tuning layer (effort, routing, fall-backs).
Five modules, one quest. The boss fight (POD-uplift embed) is in Part C; Black Belt is awarded when all three Parts close.
The five modules
Section titled “The five modules”| § | Chapter | Pillar | Time |
|---|---|---|---|
| B.7 | Progressive disclosure — skills that stay small | Context | 25 min |
| B.8 | Memory systems — auto-memory, session state, long-running agents | Context | 45 min |
| B.9 | Prompt evals — A/B, regression, golden sets | Prompt | 45 min |
| B.10 | Cost attribution + observability at team + org scale | Harness | 30 min |
| B.11 | Effort settings, model routing, fall-backs | Harness | 30 min |
Walk them in order on a first read. B.7 anchors the Part B voice; B.8 and B.9 are the depth pair (memory and evaluation); B.10 and B.11 are the scaling pair (observability and routing).
Quest B-2
Section titled “Quest B-2”Quest B-2 — Component contribution or full-stack feature. Either submit a Blade component via the contribution pipeline, or ship a feature involving a backend change you own end-to-end.
What you can say after Part B
Section titled “What you can say after Part B”“I keep skills small, evaluate them empirically, instrument cost and observability at scale, and tune the per-call layer — and I have shipped either a Blade contribution or a full-stack feature to prove cross-layer ownership.”
What comes next
Section titled “What comes next”Part C — Shape the Org: covers running office hours, embedded sprints, AI RFCs, API Council contributions, and skill / plugin governance. Closes the belt with Boss Fight B-B (POD-uplift embed) and the Black Belt badge.
Previous: ← Quest B-1 · Next: → B.7 Progressive disclosure