Simulation: Treatment participant — ForgeCraft governance
Behavior: spec authoring conversation FIRST, then ForgeCraft setup, then implementation from roadmap.
The key variable: spec quality drives everything downstream.
Note: this is also the condition JC will run manually for the treatment dry run.
— PHASE 1: Domain understanding (no code) —
Read README.md and docs/spec.md.
Before we write any code, let’s make sure we understand the domain deeply.
Write your answers to these questions to docs/DOMAIN-NOTES.md:
- What is a tanda in real life? What problem does this API solve?
- What can go wrong? List every way a user could lose money or be cheated if the rules aren’t enforced.
- What are the most ambiguous business rules — the ones where you’re not sure exactly what the code should do?
- What edge cases in the lifecycle (forming → active → completed/cancelled) are risky?
- What would make a good acceptance test for this system? (think from a user’s perspective, not a developer’s)
Do not write any implementation code yet.
— PHASE 2: Generate the spec documents —
Based on your domain analysis, write these documents:
docs/PRD.md — Product Requirements Document:
- Problem statement
- User stories (as a user I want to… so that…)
- Acceptance criteria for each user story
- Out of scope items
docs/use-cases.md — Use Cases:
- One use case per major flow (create tanda, join tanda, start tanda, record contribution, advance round)
- Each use case: actor, preconditions, main flow, alternative flows, postconditions
docs/TechSpec.md — Technical Specification:
- Data model with all fields and constraints
- API contract (all endpoints, request/response shapes, error codes)
- Business rule implementations (how each rule maps to code)
- Authentication mechanism
These documents are the source of truth. The code will follow from them.
— PHASE 3: ForgeCraft setup —
Use the forgecraft MCP tool to run setup_project on this project directory. Answer any questions it asks using the documents you just wrote as the source of truth.
Once setup_project completes, read CLAUDE.md carefully. Run: npx forgecraft-mcp audit . Fix any issues flagged before writing any implementation code.
— PHASE 4: Implement from the roadmap —
Follow the Greenfield Setup workflow from CLAUDE.md.
Scaffold the architecture layers first (src/domain/, src/services/, src/ports/, src/adapters/, src/api/). Commit the scaffold. Run forgecraft audit. Target 0 failing checks before writing any feature code.
Implement one domain at a time, using your PRD and use-cases as the spec:
- Users (create, get by token, list)
- Tanda lifecycle (create, join, start, cancel, get, list, participants)
- Contributions and rounds (record, round summary, advance, history)
After each domain: run forgecraft audit. Fix violations before moving on.
Write Vitest tests for every use case in docs/use-cases.md. Tests should verify the acceptance criteria from docs/PRD.md. Run npm test before finishing. All tests must pass.
Commit with conventional commit messages at each phase boundary.