What problem does it solve?
Manually writing tests for AI-generated or existing code and then chasing down every failure is slow and error-prone. This Skill automates the full loop: it generates progressive test layers (L0 static analysis through L3 E2E), executes them, diagnoses failures with CLI models, and applies fixes iteratively until the pass rate reaches 95%.
Core Features & Use Cases
- Two-Phase Pipeline: Phase 1 creates a test session, gathers context, analyzes requirements with Gemini, and generates task JSONs; Phase 2 runs iterative test-fix cycles with a mandatory user confirmation gate in between.
- Adaptive Strategy Engine: Automatically selects conservative, aggressive, or surgical fix strategies based on iteration count, pass rate trends, and regression detection.
- AI Code Issue Detection: Scans for hallucinated imports, placeholder code, and mock leakage common in AI-generated code, with a CLI fallback chain (Gemini → Qwen → Codex).
- Use Case: After completing an implementation session for a user authentication module, run the Skill in Session Mode to generate L0-L3 tests, then let it iterate fixes automatically until the suite passes at 95% or higher.
Quick Start
Run the workflow-test-fix-cycle skill on session WFS-user-auth-v2 to generate tests and iteratively fix failures until the pass rate reaches 95 percent.
