Refactor: create-expert delegation chain is 4 levels deep

[FL4TLiN3]

Description

The create-expert system has a 4-level delegation chain which may introduce latency and complexity:

create-expert (level 1)
└── integration-manager (level 2)
    └── functional-manager (level 3)
        └── expert-tester (level 4)

Current Architecture

create-expert
├── property-extractor (level 2)
├── ecosystem-builder (level 2)
├── integration-manager (level 2)
│   ├── functional-manager (level 3)
│   │   └── expert-tester (level 4)
│   └── usability-manager (level 3)
│       └── expert-tester (level 4)
└── report-generator (level 2)

Potential Issues

1. Latency: Each delegation level adds LLM round trips
2. Context loss: Each delegation creates a fresh context, making it harder to maintain coherent understanding
3. Debugging difficulty: Failures 4 levels deep are harder to trace

Consideration

Per docs/understanding-perstack/experts.md:

Context is never shared between Experts. The delegate receives only the query — no message history, no parent context.

At 4 levels deep, the original requirements from create-expert are being summarized multiple times, increasing risk of information loss.

Possible Simplification

Could integration-manager directly delegate to expert-tester?

create-expert
├── property-extractor
├── ecosystem-builder
├── integration-manager
│   └── expert-tester (handles functional + usability)
└── report-generator

Or merge functional-manager and usability-manager with expert-tester?

Affected Areas

• apps/create-expert/src/lib/create-expert-toml.ts (delegation structure)

Acceptance Criteria

• Evaluate if 4-level depth is necessary
• Either simplify the chain or document why the depth is required

[github-actions]

🤖 Processing...

---