test-suite

Test Suite — subcommand router

/test-suite <subcommand>. Pick the operation, read its reference, follow it. Default subcommand is create. Most real tasks chain two or three.

Subcommands

/test-suite …	You need to…	Reference
create (default)	Design a minimal-but-covering strategy for a TODO/task, split into unit / integration / manual tiers via pairwise.	references/create.md
write	Enumerate scenarios + a coverage matrix before implementation (.md index + Gherkin features).	references/write.md
verify	Audit, score, or validate an existing test set for missed cases before merge — Ready / Not Ready verdict.	references/verify.md
case-design	Derive cases systematically — equivalence partitioning, boundary values, decision tables, state transition, state-combination.	references/case-design.md
bdd	Shape integration/e2e cases as Cucumber Given/When/Then.	references/bdd.md (+ references/gherkin-guide.md)
tdd	Drive a feature or bug-fix spec-before-code with Red-Green-Refactor.	references/tdd.md (+ references/bdd-best-practices.md)
harness	Test a harness plugin in isolation — MCP server, unit tests, typecheck, Claude plugin loading (tmux panes).	references/harness.md
review	Run the verify phase — present results, let the user review an implementation against acceptance criteria.	references/review.md

Supporting references (not direct subcommands): references/gherkin-guide.md and references/bdd-best-practices.md — backing docs for bdd and tdd.

How they combine

• case-design is the technique toolbox — feeds the case lists of create and write.
• create sizes the strategy across cost tiers (unit/integration/manual) with pairwise; write turns it into concrete artifacts (behavioral → Gherkin, unit → table — never both).
• bdd / tdd govern how behavioral cases read and how implementation is driven.
• verify audits the finished set and returns a Ready / Not Ready verdict.
• harness and review are standalone — execution (plugin smoke tests) and review (user sign-off), not case design.

Typical chain for a TODO: case-design (derive) → create (tier + pairwise) → write (write artifacts) → verify (audit).

Output shape

The canonical strategy document is in examples/strategy-auth-refresh.md. Per-TODO output goes into the ## Autotest and ## Manual test sections of <notes-dir>/todos/TODO-N.md (follow the code skill's todo subcommand); task-wide output goes to <notes-dir>/test-strategy.md.