Test Design Orchestrator

test-design-orchestrator is an AgentSkill for turning requirements into structured, traceable software test artifacts. It is built as a composite skill: the root skill chooses the best-fit black-box test design technique, routes execution to a technique-specific subskill, and optionally formats the result for downstream tooling.

This repository is intentionally split into two layers:

• the installable skill itself: SKILL.md, subskill folders, references/, assets/templates/, agents/openai.yaml
• repository-facing support material: README.md, examples/, evals/, and validation scripts

What This Skill Does

• Selects an appropriate test design technique for a given requirement shape.
• Generates test artifacts using boundary value analysis, equivalence partitioning, decision tables, classification trees with n-wise reduction, state transitions, acceptance criteria, or use case analysis.
• Preserves traceability, assumptions, and coverage notes.
• Formats output for review-ready markdown, BDD feature files, Xray-compatible Gherkin feature bundles, Zephyr Scale CSV, or TestLink-oriented import workflows.

What This Skill Does Not Do

• It does not replace domain requirements analysis. Missing business rules are surfaced, not invented.
• It does not provide shared-memory infrastructure. Cross-agent memory belongs in a separate shared-memory skill.
• It does not claim support for every ALM tool beyond the bundled targets and references.

Repository Layout

.
|-- SKILL.md
|-- agents/
|   `-- openai.yaml
|-- acceptance-criteria-to-test-cases/
|-- boundary-value-analysis/
|-- classification-tree-nwise/
|-- decision-table/
|-- equivalence-partitioning/
|-- state-transition/
|-- technique-selector/
|-- test-case-formatter/
|-- use-case-testing/
|-- references/
|-- assets/templates/
|-- examples/
|-- evals/
`-- scripts/

Supported Workflow

1. Normalize the request into inputs, rules, actors, states, and desired outputs.
2. Choose one primary test design technique with a defensible rationale.
3. Generate the test artifacts using the matching subskill.
4. Format only when the user asks for a specific target representation.
5. Return assumptions, traceability, and residual risk with the artifact unless the user requested artifact-only output.

Installation

Install from GitHub with npx skills:

npx skills add <owner>/<repo> --skill test-design-orchestrator

If your installer supports direct GitHub URLs, this form is also commonly used:

npx skills add https://github.com/<owner>/<repo> --skill test-design-orchestrator

After installation, restart Codex so it reloads the newly installed skill.

For a manual install, place this folder in your skill directory and publish it under a lowercase hyphen-case folder name such as test-design-orchestrator.

The required installable files are:

• SKILL.md
• subskill SKILL.md files
• agents/openai.yaml

Everything else is supportive but strongly recommended for maintainability.

Inputs and Outputs

Useful inputs include:

• raw requirements
• business rules
• user stories and acceptance criteria
• use cases
• lifecycle or state descriptions
• a requested export target such as markdown, BDD, Xray Gherkin, Zephyr Scale, or TestLink

Typical outputs include:

• technique recommendation with rationale
• partition tables and representative values
• decision tables and optimized rules
• transition paths and invalid-transition tests
• scenario lists and detailed test cases
• import-oriented formatted artifacts

Memory Model

Runtime memory:

• ephemeral only
• used for requirement normalization, assumptions, traceability IDs, and current output format

Project-local persistent memory:

• optional
• only appropriate when the user explicitly wants a reusable test-design brief saved in their project

Shared memory:

• intentionally excluded from this repository
• integrate an external shared-memory skill if cross-agent reuse is required

Validation and Evaluation

Run the repository checks:

python scripts/validate-skill-repo.py
python scripts/format-validator.py bdd examples/checkout-feature.feature
python scripts/package-xray-features.py examples/xray-checkout.feature --output dist/xray-features.zip

Use the evaluation fixtures:

• evals/trigger-queries.json for description triggering checks
• evals/technique-selection-cases.json for manual or agent-assisted forward testing

Use the prompt examples in examples/ to sanity-check the end-to-end workflow.

Optional Integrations

• assets/templates/zephyr-scale.csv.j2 for Zephyr Scale CSV generation
• assets/templates/xray-gherkin.feature.j2 plus scripts/package-xray-features.py for Xray Gherkin feature import bundles
• references/testlink-import-file-formats.pdf for TestLink import guidance
• assets/templates/bdd-feature.j2 plus the BDD protocol references for feature-file output

These integrations are supported only to the extent that the bundled templates and references cover them. If a target tool requires fields that are not available in the input, the skill should stop and ask for them.

Contributing

Keep changes tightly scoped and auditable:

• improve trigger quality by updating SKILL.md and evals/trigger-queries.json together
• keep technique guidance in references/ concise and technique-specific
• keep renderable output contracts in assets/templates/
• add or update example prompts whenever the workflow changes materially

License

This repository is released under the MIT License. See LICENSE.

Publishing Notes

Before publishing to GitHub or a skill registry:

• rename the repository folder to lowercase hyphen-case
• run python scripts/validate-skill-repo.py
• optionally run the upstream skill quick validator against the root folder