matrix_skill

Overview

This skill is a domain-agnostic planning agent that controls combinatorial explosion by selecting minimal coverage sets from any multi-dimensional axes × values input. It designs prioritized execution plans (tests, deploys, UX checks, risk assessments, experiments, compatibility matrices) and hands off structured plans to execution agents. Matrix never executes code; it produces optimized, traceable plans and coverage visualizations.

How this skill works

Matrix parses inputs in natural language, YAML, JSON, or table form to extract axes, values, and constraints. It enumerates the Cartesian space, reports total combination counts, then applies an appropriate optimization method (Pairwise, Orthogonal Array, or custom constraint-aware algorithms) to choose a minimal coverage set. The agent scores and ranks combinations by Risk × Impact × Cost, emits a handoff-ready execution plan for downstream agents, and provides coverage metrics and back-mapping placeholders for post-run results.

When to use it

• You have three or more axes and cannot run all combinations due to cost or time limits.
• You need a prioritized, minimal set of scenarios for testing, deployment, UX validation, risk assessment, or experiments.
• Combinatorial explosion prevents clear decision-making about what to run next.
• Downstream execution will be handled by specialized agents and you need a handoffable plan.
• You must guarantee pairwise or higher-order coverage while minimizing runs.

Best practices

• Always supply explicit axis names and values; use YAML if you expect reuse or automation.
• State constraints and cost/time limits up front so Matrix can pick the correct optimization strategy.
• When axes ≥ 6 or priorities are unknown, confirm priorities before optimization.
• Require a handoff format target (Voyager, Siege, Echo, Experiment, Scaffold, Triage, Canvas) in the request.
• Review the coverage rate and the chosen method rationale before executing downstream plans.

Example use cases

• Generate a browser × OS × auth-state minimal E2E test plan and hand off to a test implementation agent.
• Compress load-test parameter space (concurrency × payload × endpoint × time) and send to a load runner.
• Design a deploy matrix across region × version × traffic-split and produce a rollout plan for deploy agents.
• Create a persona × device × scenario UX test set for qualitative researchers or UX automation.
• Map library × version × runtime compatibility cases and hand off prioritized checks to build agents.