team-testing_skill

Overview

This skill implements a unified testing team workflow that routes work to role-specific logic via a --role argument. It coordinates strategist, generator, executor, analyst, and coordinator behaviors using a shared session directory, shared-memory, and a message bus. The design enforces role isolation, progressive coverage pipelines, and a generator-critic loop to iteratively improve tests until coverage targets are met.

How this skill works

On invocation the skill parses --role and --team arguments, validates the role, and reads the corresponding role file to execute role-scoped logic. Workers operate on tasks filtered by task-prefix and owner, update task status, log via the message bus, and read/write only their fields in shared-memory.json. Pipelines (targeted, standard, comprehensive) drive role handoffs and the generator↔executor loop enforces test revision when coverage falls short.

When to use it

• Automate multi-role test workflows in a CI-like environment
• Enforce role separation and audit-friendly logging for testing activities
• Drive progressive test coverage across unit, integration, and E2E layers
• Coordinate multi-agent or multi-worker test pipelines with shared state
• Implement automated generator-critic loops to raise coverage to targets

Best practices

• Always supply --role and --team when invoking the skill; role is required
• Respect output tagging conventions (e.g., prefix summaries with [role]) for message bus tracking
• Only read/write your role-specific fields in shared-memory.json to avoid conflicts
• Use coordinator for pipeline selection and quality gates; workers must not bypass it
• Limit GC loop iterations (default 3) before accepting current coverage with a warning

Example use cases

• A strategist analyzes a git diff and creates STRATEGY-* tasks that select L1 unit tests for a small change
• A generator produces TESTGEN-* artifacts per layer and commits tests under the session tests/ folder
• An executor runs TESTRUN-* jobs, collects coverage, and triggers TESTGEN-fix if targets miss
• An analyst consumes execution results to generate a quality report and defect-pattern history in analysis/
• Coordinator spawns team tasks, monitors the pipeline, and enforces quality gates across layers