TDD for AI Agents
Reusable test-driven development workflow for AI coding agents
The always-on skill that enforces strict TDD discipline, requirements documentation, and tiny test-verified steps — the orchestrator that makes Kindly Web Search and Lad Code Review work together as a unified AI coding suite.
"Fantastic setup. Improved my Claude Code generation quality by 15–20% for our company’s commercial projects."
Part of Our AI Coding Suite
TDD for AI Agents orchestrates our complete open-source toolkit. It calls Kindly to verify assumptions and Lad to review designs and code — ensuring all three tools work together seamlessly.
REQUIREMENTS.md
Every task begins with a structured requirements document. This eliminates ambiguity and gives the AI agent a clear, testable contract to work against.
1As Is
Current behavior and state — what exists today before any changes.
2To Be
Desired behavior after the change — the clear target state.
3Requirements
Numbered functional requirements with explicit acceptance criteria.
4Testing Plan
Test strategy and cases following TDD best practices.
5Implementation Plan
Smallest sequential changes, each with its own test verification.
6System Design Review
Reviewed by Lad Code Review before implementation begins.
Workflow
A strict six-step process that turns vague requirements into production-ready, tested code.
1.Investigate
Inspect repository structure, code paths, current behavior, and existing tests. Run minimal commands to confirm the current state.
2.Clarify & Write Requirements
Ask clarifying questions, propose testable requirements, get user confirmation, then write REQUIREMENTS.md with the full mandated structure.
3.Validate Assumptions
Use Kindly Web Search to verify API signatures, version behavior, breaking changes, and deprecations. Update requirements if assumptions change.
4.Design Review with Lad
Submit the system design in REQUIREMENTS.md to Lad's system_design_review tool. Iterate until feedback converges.
5.Implement with TDD
Follow the Implementation Plan in order. Write tests first, implement features second, verify tests pass. Keep diffs minimal and focused.
6.Code Review & Verify
Use Lad's code_review tool after each change. Implement worthy suggestions. Iterate until Lad stops providing new feedback.
Non-Negotiables
These rules are always enforced. No shortcuts, no exceptions — unless the user explicitly opts out.
One Suite, Three Tools
TDD for AI Agents is the orchestrator that ties our open-source AI coding tools into a complete development workflow. Each tool has a specific role.
Kindly Web Search
Verifies API docs, finds solutions, retrieves full technical context in a single call.
Lad Code Review
Provides a fresh pair of AI eyes — catches coherence-driven mistakes the coding agent misses.
TDD Skill (Orchestrator)
Enforces the workflow, writes requirements, coordinates research and review steps.
Supported AI Coding Assistants
Ready to bring TDD discipline to your AI coding workflow?
Deploy the full suite in your private cloud or hire the engineers who built it.