Turn system and release evidence into a risk-based test strategy with traceable coverage, environments, data, execution gates, and a clear readiness decision.
A test count is not a strategy. Neither is an 80% coverage target or a pyramid copied into a planning document. The hard part is deciding which failures matter, what evidence can expose them, where that evidence is trustworthy, and what must be true before execution or release can move forward.
Test Strategy builds that decision system. It starts with the product and architecture evidence, maps critical flows and failure consequences, then binds each material risk to a test level, oracle, environment, data set, execution trigger, owner, and evidence artifact. It covers contract boundaries, migrations, rollback, offline and replicated state, flaky tests, sensitive data, production-only uncertainty, and controlled strategy updates after incidents or architecture changes.
Use it for a new product, a risky release, an incomplete inherited test plan, or a strategy delta after the system changes. The output separates observed facts from proposals and open questions. Planned, implemented, executed, and passing coverage stay distinct, so a green-looking document cannot quietly stand in for results.
Version 1.0.0 was reviewed on 2026-07-17 and evaluated through Codex CLI with gpt-5.6-sol in 16 fresh read-only contexts. The candidate scored 100 across four cases, compared with 86.25 for the same agent without the skill and 93.75 for the exact free upstream. The cases covered a payment release with migration and exactly-once business risks, sensitive offline sync with missing policy and environment evidence, a blocked canonical handoff, and an approved canonical handoff. This evaluation measures the supplied cases. It does not prove a particular system is tested, a release is safe, an environment matches production, or an organization has met legal or compliance duties. The reviewed package SHA-256 is 5be39a1f728951c34e98548417b212d6de4185b35496f257530681e809b1a9bf.
The package includes a read-only checker for an optional canonical JSON strategy record. Passing it confirms structure, references, risk arithmetic, coverage bindings, gate state, and verdict precedence. It cannot verify source truth, run a test, inspect an environment, or approve a release. Maintenance and install support are handled by the platform-operated publisher for this listing.
Input
Build the test strategy for checkout v18. We changed idempotency, added webhook retries, and have no migration or rollback rehearsal. Staging uses one worker and a fake payment provider.
Output
A source and authority register; critical-flow and risk model; risk-to-coverage matrix for idempotency, contracts, migration, rollback, concurrency, tenant isolation, and reconciliation; environment and data gaps; execution triggers; release and canary gates; owners; and a blocked verdict until critical evidence exists.
Input
Review this health-journal offline-sync plan. It says 80% line coverage, ten unit tests, and manual checks on two phones. The conflict policy, API contract, privacy rules, and rollback plan are missing.
Provide the system or change in scope, target revision, critical user and business outcomes, architecture and interface evidence, incident history, existing tests, environments, data constraints, delivery cadence, rollback path, observability, owners, and permission boundary. Ask for a new strategy, a review of an inherited plan, a release-specific strategy, or a controlled delta after the system changes. The skill keeps missing evidence and authority visible instead of inventing coverage or results.
The product, component, release, migration, or program to evaluate; target revision; supported platforms; critical outcomes; and the decision the strategy must support.
Current requirements, diagrams, interfaces, data flows, dependencies, incidents, operational signals, and existing test or delivery artifacts with revisions or observation times.
Available environments, test data, tools, people, time, budget, release cadence, rollback path, observability, and known production differences.
Strategy reviewer, gate and release decision owners, and whether the agent may only plan or may also inspect, create, execute, mutate configuration, or access production.
SKILL.md; risk, coverage, environment, data, execution, flake, gate, and maintenance method; concise and canonical JSON output contract; read-only strategy-record checker; Agent Skills interface metadata
No reviews yet.
Output
An evidence receipt; offline and replicated-state failure model; prioritized privacy, integrity, deletion, migration, attachment, and recovery risks; property, contract, persistence, device, accessibility, and network-failure coverage; controlled test-data requirements; observable gates; and a blocked verdict with exact recovery questions.
Input
Create the canonical strategy record for this approved tenant-isolation contract test, including the source, flow, risk, environment, synthetic data, coverage, release-candidate trigger, passed gate, authority, and readiness scope.
Output
A schema-bound JSON record with resolvable IDs, correct risk arithmetic, explicit permission and evidence states, approved high-risk coverage, a passed gate, no hidden blocker, and a scoped ready-for-execution verdict that the bundled checker can validate.
Creator
Nnuravo43
