PRD — Capability-Architect

Use it for

Ambient-intelligence walkthrough — capability-architect as compiler

The prompt

**Type:** PRD (orchestrator to be built) | **Status:** Extracted — pending design review | **Priority:** High
---
## Problem Statement
Asking an AI to "build me an agent / writing team" produces a single mega-prompt that's hard to audit, hard to optimize, and not reusable. There's no disciplined process that decomposes a workflow into the right execution shape (inline / fork / spawn / parallel), routes model + effort per step, defines what each child returns, and packages the result as an inheritable skill bundle.
> *"This is different from 'make me an agent.' It is a disciplined design process for capability creation."* — Lou
---
## What It Should Do
A compiler, not a "make me an agent" prompt. Takes a workflow, problem, or existing pipeline and walks a fixed compile path:
1. Intake — state the job in one sentence; ask clarifying questions if it can't
2. DAG — identify which artifacts depend on which
3. Classification — label each step code / inference / hybrid
4. Bundling — group steps whose intermediate outputs are only used internally
5. Execution — decide inline / fork / spawn / parallel per step
6. Routing — assign model + effort + rationale per component (least-excessive inference)
7. Contract — define what each child returns (JSON envelope + artifact path)
8. Generation — write the actual skill bundle + local assets
9. Evaluation — define trigger fire / no-fire cases
10. Install — make the bundle inheritable (publish as a plugin)
**Output:** An inheritable skill bundle that drops into the ambient library.
---
## Fork vs Spawn Decision Rule
- Yes → fork (inherits parent context)
- No → spawn (starts cold; use when inherited context would contaminate the work)
---
## Known Gotchas
- Forks must be able to run with the context they're given — verify context sufficiency at design time
- Don't default everything to the strongest model at high effort — assign least-excessive inference
- Skill-creator inflates front-matter descriptions to the 1024-char max — cap descriptions tight when generating
---
## Composition
Assumes `inference-router` is available. Overlaps with skill-creator (generation step) and build-from-prd.
**Source:** Lou — June 11, 2026 Mastermind

Worked example

No worked example recorded yet.