Software Factory Stack Boundaries
Give every component one clear responsibility and replacement boundary.
A focused view of boundaries, contracts, state, authority, failure paths, and tradeoffs drawn from this chapter.
Reconstruct and defend this chapter’s architecture.
Reconstruct the architecture, name each boundary, and defend the tradeoffs.
Open the source exercise
Draw a composed stack in which the organization owns the control plane and development environment, buys an inner harness, and uses managed compute. Mark every identity, contract, event stream, credential, policy decision, fallback, and failure owner. Then replace the inner harness without changing acceptance authority.
4. Tradeoffs and alternatives
A vertically integrated stack reduces integration work and can deliver a better initial experience. It also concentrates policy, telemetry, execution, and environment assumptions in one provider. A composed stack preserves choice but moves compatibility testing and failure ownership to the adopter.
The inner/outer harness split is a conceptual boundary, not a requirement for two deployable services. A small system may implement both in one process while retaining separate contracts. Conversely, calling every wrapper an outer harness can create abstraction for its own sake. Introduce the boundary only when it clarifies replacement, authority, testing, or operations.
5. Current Mission Control Implementation
At study commit
d902fae,
Mission Control has a clear control-plane and execution-plane doctrine. Convex
owns durable authority and state, the Hono service hosts orchestration and
provider boundaries, and provider-neutral harness lifecycle contracts describe
execution through capability manifests and structured results.
The implementation has versioned agent records, skill discovery and linting, model routes, context packages, sandbox profiles, and evaluation mechanisms. Those capabilities resemble parts of an Agent Factory, but the studied product does not establish one canonical Agent Factory boundary with a unified capability publication, admission, compatibility, deprecation, and revocation lifecycle. Exact skill-version binding in the execution manifest also remained incomplete.
The production execution path remained blocked by operator configuration at the studied commit. The architecture and local qualification therefore support the boundary model; they do not prove a live, fleet-scale composed stack.
Review this chapter.
Challenge a claim, boundary, missing failure mode, unclear term, or unsupported evidence statement.
- Claim
- Boundary
- Failure
- Evidence