Autonomous TDD Agent

SkillSecurity

Lets your agent write working Python code that passes your existing pytest tests, following a six-step cycle.

Available today. Use it from your connected AI after setup.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the Autonomous TDD Agent skill

About this skill

Autonomous Test-Driven Development Agent workflow for implementing features without human intervention.

What this skill tells your AI

The instructions your AI receives, as published by shadd0wtaka/zen-ai-pentest in .skills/autonomous-tdd-agent/SKILL.md and read by ahel’s review.

Execute Test-Driven Development autonomously without asking for clarification. Make reasonable assumptions, document them, and proceed through the implementation loop.

The 6-Step Autonomous Loop

Step 1: Analyze Failing Tests

  • Read the test file completely
  • Identify all test classes and methods
  • Understand what the tests expect (inputs, outputs, side effects)
  • Note required classes, functions, and their signatures
  • Document assumptions in comments

Step 2: Initial Implementation

  • Create the minimal implementation to satisfy the tests
  • Focus on correctness over elegance
  • Include proper imports and exports
  • Add basic error handling

Step 3: Predict Test Results

  • Before running, predict which tests will pass/fail
  • Document your prediction as a comment
  • Run the actual tests
  • Compare prediction vs reality

Step 4: Iterate Until Green

  • Fix failing tests one by one
  • Document each fix in code comments
  • Never skip a failing test
  • If stuck, try alternative approaches

Step 5: Refactor

  • Once all tests pass, improve code quality
  • Extract functions, add type hints, improve naming
  • Ensure no tests break during refactoring
  • Remove duplication

Step 6: Summarize

  • Provide summary of changes across iterations
  • Document key decisions and assumptions
  • Note test coverage achieved

Workflow Pattern

# 1. READ TESTS
# Read and understand all test requirements

# 2. IMPLEMENT
# Write initial implementation

# 3. TEST & ITERATE
while failing_tests > 0:
    run_tests()
    fix_failures()
    document_fixes()

# 4. REFACTOR
improve_code_quality()
verify_tests_still_pass()

# 5. SUMMARIZE
report_changes()

Rules

  1. Never ask for clarification - Make reasonable assumptions and document them
  2. Implement both approaches - If two valid approaches exist, implement both as branches
  3. Minimal changes - Only change what's necessary to make tests pass
  4. Document everything - Every assumption, every fix, every iteration
  5. Test everything - All code paths must be covered

Example Usage

Feature: Implement a rate limiter
Test file: tests/unit/test_rate_limiter.py

[Execute 6-step loop autonomously]

Output Format

After completion, provide:

  1. Final implementation code
  2. Iteration summary table
  3. Test results
  4. Key assumptions made

See references/WORKFLOW_EXAMPLES.md for detailed examples.

Signals

GitHub stars
469
Forks
81
Last commit
Sep 2026
Advanced
Item type
skill
Key
autonomous-tdd-agent
Source
github.com/shadd0wtaka/zen-ai-pentest