parcadei/tdd-migrate
TDD workflow for migrations - orchestrate agents, zero main context growth
npx skills add https://github.com/parcadei/Continuous-Claude-v3 --skill tdd-migrateOrchestrate TDD migrations with agents doing all work. Main context stays clean.
/tdd-migrate <source_path> <target_path> --pattern <reference> --items "item1,item2,item3"
source_path: Path to analyze (existing code)target_path: Where to create new codepattern: Reference file/pattern to followitems: Comma-separated list of things to createPhase 0: YAML TODO List
│
▼
Phase 1: TLDR Analysis ─────────────────┐
│ │
▼ │ Parallel scouts
Phase 2: Write Failing Tests ───────────┤ per item
│ │
▼ │
Phase 3: Implement (minimal) ───────────┤
│ │
▼ │
Phase 4: Build + Pass Tests ────────────┘
│
▼
Phase 5: QLTY Check ────────────────────┐
│ │ Parallel
Phase 6: Review Agent Validates ────────┘
│
▼
Phase 7: TLDR Diff (new vs reference)
│
▼
Phase 8: Fix Issues (if any)
│
▼
Complete
| Task | Agent |
|------|-------|
| Explore/analyze | scout |
| Write tests + implement | kraken |
| Quick fixes | spark |
| Run tests/validate | validator |
| Code review | critic |
Write a YAML plan file to thoughts/shared/plans/<name>-tdd.yaml:
---
title: <Migration Name>
date: <today>
type: implementation-plan
approach: TDD (test → build → pass → review)
items:
- name: item1
file: <target_path>/item1.ts
test: <target_path>/__tests__/item1.test.ts
deps: []
- name: item2
# ...
reference: <pattern_file>
workflow:
per_item:
1: Write failing test
2: Implement minimal
3: Build
4: Pass test
5: QLTY check
6: Review
final:
7: Integration test
8: TLDR diff
Task (scout): Analyze <source_path> with TLDR
Task (scout): Analyze <pattern> to understand structure
Task (scout): Read migration handoff if exists
For each item, launch ONE kraken that does full TDD:
Task (kraken): Implement <item> using TDD workflow
1. Read pattern file
2. Write failing test
3. Implement
4. Run: bun test <test_file>
5. Run: qlty check <impl_file>
Task (critic): Review all new files against pattern
Task (validator): Run full test suite
Task (validator): QLTY check all files
If critic/validator found issues:
Task (spark): Fix <specific issue>
Task (validator): Re-validate
Task (validator): TLDR diff new files vs reference
- tldr structure <new_file> --lang <lang>
- tldr structure <reference> --lang <lang>
- Compare patterns
Update ledger with completed work.
/tdd-migrate /Users/cosimo/Documents/rigg/src/sdk/providers \
/Users/cosimo/Documents/rigg/src/sdk/providers \
--pattern lmstudio.ts \
--items "xai,cerebras,togetherai,deepinfra,perplexity"
Resulted in:
| Bad | Good |
|-----|------|
| Read files in main context | Launch scout agent |
| Write code in main context | Launch kraken/spark agent |
| Run tests in main context | Launch validator agent |
| Skip review | Always launch critic |
| Sequential items | Parallel krakens |
| Fix in main context | Launch spark |
Explore <path> to understand:
1. Structure/patterns
2. Interfaces/types
3. Dependencies
Return actionable summary for implementation.
Implement <item> using TDD:
1. Read <pattern> for structure
2. Write failing test to <test_path>
3. Implement minimal to <impl_path>
4. Run: <test_command>
5. Run: qlty check <impl_path>
Report: status, issues, files created.
Review <files> against <pattern>:
1. Pattern compliance
2. Type safety
3. Missing registrations
4. Security issues
DO NOT edit. Report issues only.
Fix <specific issue>:
1. Read <file>
2. Make minimal edit
3. Verify fix
Validate <files>:
1. Run <test_command>
2. Run qlty check
3. Report pass/fail/issues
Take parcadei/tdd-migrate from the repository into ~/.claude/skills for personal
use, or into .claude/skills inside a project.
The agent identifies a skill by the name field in its header. Two skills with the
same name cannot sit side by side — one of them will be ignored.