dmitryprg-ai/development

Development Protocol

Pre-Action (MANDATORY before coding)

1. Duplicate Check

## DUPLICATE CHECK
**Target:** [what I'm creating]
**Search:** semantic/grep/file -> [found/not found]
**Decision:** CREATE NEW / EXTEND EXISTING / ASK USER

2. JTBD Analysis (for user-facing features)

## JTBD ANALYSIS
**Job Story:** When [context], user wants [action], to get [result] and [benefit]
**Questions:**
1. What task are they solving? -> [answer]
2. What's blocking them? -> [answer]
3. What change will help? -> [answer]

3. Find Working Example

For UI: find a working pattern in the project first. Copy the pattern, then adapt. Do NOT invent CSS classes without checking they work.

4. Control File Size

Plan logic across multiple files. Refer to standard-file-size-limits-agent.mdc rule.

Plan

## DEVELOPMENT PLAN
**Task:** [what to do]
**Expected Output:** [artifacts]
**Test Cases:** [input -> expected output]

Execute (TDD approach)

Follow the tdd-workflow skill for full TDD process (Red → Green → Refactor):

1. Fill test cases table BEFORE code (Phase 0)
2. Write failing tests (Phase 1 — RED)
3. Write minimal code to pass (Phase 2 — GREEN)
4. Refactor (Phase 3)
5. Restart affected services and verify

Verify

• [ ] 0 linter errors
• [ ] Tests pass
• [ ] JTBD: Job implemented, UI about benefits
• [ ] Each file opened and verified
• [ ] Services restarted if needed

Common Pitfalls

• UI Components: Find working example in project first. Don't use CSS classes without checking they work.
• Type Consistency: Database types may differ from JSON (bigint -> string). Check actual types.
• Backend Modules: Follow pattern: types -> repository -> service -> routes
• External APIs: List/Enum fields may contain IDs, not text. Need mapping.
• Entry Points: Use high-level orchestrator services, not low-level methods.

Additional Resources

For prompt preparation methodology, see PREPARE-PROMPT.md.