ddd

Prompt file imported from acailic/agent_debugger (.claude/commands/ddd.md). Fill in {{arguments}} before use. Copyright stays with the author.

Domain-Driven Design Assistant

Arguments: {{arguments}}

Determine Mode

Decide the mode from the arguments above:

• If the argument is empty, or looks like a file/module path such as agent_debugger_sdk/..., api/..., frontend/..., or tests/..., use Review Mode.
• If the argument looks like a domain concept such as session replay ranking, policy violation provenance, or checkpoint restore value, use Model Mode.

Review Mode

Analyze the project, or the target path, through a DDD lens. Ground the review in the real repo structure rather than generic DDD advice.

1. Map Bounded Contexts

Use these repo-specific contexts:

• Tracing Core: agent_debugger_sdk/core/, agent_debugger_sdk/checkpoints/, agent_debugger_sdk/config.py, agent_debugger_sdk/transport.py, agent_debugger_sdk/pricing.py
• Instrumentation Edge: agent_debugger_sdk/adapters/, agent_debugger_sdk/auto_patch/
• Server / Delivery: api/, plus collector/, storage/, auth/, and redaction/ when they participate in the target flow
• Presentation / Inspection UI: frontend/src/
• Contract Boundary: api/schemas.py, frontend/src/types/index.ts, frontend/src/api/client.ts

If the target touches one context, still note the adjacent contexts it depends on.

2. Entities, Value Objects, and Aggregates

For the contexts you inspect:

• Identify likely entities such as sessions, traces, checkpoints, alerts, or replay analyses.
• Identify likely value objects such as rankings, summaries, policy parameters, or tool call payloads.
• Identify aggregate roots and consistency boundaries.
• Flag places where persistence shape, transport shape, and domain shape are being conflated.

3. Domain Logic Leakage

Look for:

• Business rules in FastAPI routes instead of services or SDK/core modules
• Server or storage details bleeding into tracing core
• Frontend components re-implementing backend/domain decisions instead of consuming a clear contract
• Serialization or API schemas owning rules that should live in domain logic

4. Boundary Health

Check whether dependencies flow cleanly:

• core/checkpoints should stay independent of API and frontend
• adapters/auto-patch may depend on core, but should isolate framework-specific models
• API should expose contracts rather than force the frontend to know SDK internals
• frontend should depend on serialized shapes, not backend implementation details

Also check whether changes to event shapes require synchronized edits across:

• api/schemas.py
• frontend/src/types/index.ts
• frontend/src/api/client.ts
• matching tests in tests/

5. Recommendations

Return a prioritized list of concrete improvements. For each one:

• state the current boundary problem
• identify the exact file(s) involved
• propose a concrete refactor
• rate impact as high, medium, or low

Model Mode

Help the user model the domain concept in the context of this repo.

1. Place the Concept

• Restate the concept in Peaky Peek terms.
• Identify whether it belongs mainly to tracing core, instrumentation edge, server delivery, or frontend inspection.
• If the concept is ambiguous, ask a clarifying question and wait.

2. Tactical DDD Design

For the concept, propose:

• Entity: identity and lifecycle
• Value Object: immutable structured data
• Aggregate: consistency boundary and root
• Domain Event: emitted or consumed events
• Repository / Storage Boundary: how it is saved or queried
• Service: behavior that does not fit cleanly on an entity

3. Integration Points

Call out how the concept affects:

• SDK event capture
• API schemas/routes/services
• frontend contract types and UI panels
• tests that should be added or updated

4. Proposed File Layout

Output a concrete file tree that matches this repo's conventions. Prefer existing directories such as:

agent_debugger_sdk/core/
agent_debugger_sdk/checkpoints/
agent_debugger_sdk/adapters/
agent_debugger_sdk/auto_patch/
api/
frontend/src/
tests/
tests/auto_patch/

Reference existing files that are close analogs when possible.