workflow-test-fix-cycle_skill

Overview

This skill implements an end-to-end test-fix workflow that generates progressive test sessions (L0–L3) and runs iterative fix cycles until a pass rate of at least 95% is achieved. It combines test generation and execution into a single autonomous pipeline that dispatches and coordinates lightweight subagents for analysis, planning, and fixes. The tool is JSON-driven and designed for CLI-first orchestration in TypeScript projects.

How this skill works

On trigger, the pipeline runs two phases: Phase 1 generates a test session and produces task JSONs and analysis artifacts; Phase 2 executes those tasks with an adaptive fix loop until the quality gate is met. Subagents are spawned for context gathering, CLI analysis, test generation, planning, and fix application; their outputs are parsed and stored under the workflow directory. The strategy engine shifts between conservative, aggressive, and surgical tactics based on iteration progress and test criticality.

When to use it

• Validate and harden AI-generated code or new implementations with automated tests.
• Run a fully autonomous test-and-fix cycle for a feature branch or completed implementation.
• Enforce quality gates (>=95% pass rate) before merging critical changes.
• Continuously iterate on flaky or partially passing test suites until stable.
• Resume interrupted test sessions without re-generating test artifacts.

Best practices

• Provide a clear feature description or a completed implementation session (WFS-*) to seed Phase 1.
• Review IMPL_PLAN.md and TODO_LIST.md produced in Phase 1 before long-running Phase 2 runs when manual oversight is needed.
• Set --max-iterations to control cost and avoid infinite loops for large repos.
• Ensure role files for subagents exist at the expected path so spawned agents read required directives first.
• Let the orchestrator run autonomously but monitor iteration summaries and fix-history.json for regressions.

Example use cases

• Generate unit, integration, and optional E2E tests for a newly delivered API endpoint and run automated repair cycles until 95%+ coverage and pass rate.
• Resume a previously created session to continue fix iterations after an interrupted CI run without regenerating tests.
• Hardening AI-completed code by detecting hallucinated imports, placeholder logic, and mock leakage via L0 static checks before running tests.
• Apply a surgical fix strategy for a small set of high-criticality failing tests while leaving stable components untouched.