Ouroboros Test Plan Author

Use this skill to turn a high-level feature, module, bug, or goal into one or more executable Markdown charters under test-plan/.

Workflow

1. Ground the plan in the repo before writing:
   • Search relevant renderer components, stores, protocol types, existing E2E specs, and existing test-plan/*.md.
   • Prefer current labels, UI names, domain language, and mock data patterns from the codebase.
2. Decide plan granularity:
   • One plan for one user-centered workflow.
   • Split broad modules into multiple charters only when the workflows have different setup, risks, or pass/fail oracles.
3. Write the Markdown plan in test-plan/<kebab-case-name>.md.
4. Keep the plan intent-based:
   • Describe what the user is trying to accomplish.
   • Avoid brittle CSS selectors, implementation details, or Playwright-style commands.
   • Include enough observable outcomes for an Agent Browser runner to judge pass/fail.
5. If the plan needs mock data, name the required scenario shape in plain language.
   • Do not invent large fixture JSON unless the user explicitly asks for it.
   • State when missing mock data should produce INCONCLUSIVE.

Required Plan Shape

# Title Case Plan Name

## Mission

One or two sentences describing the user workflow and why it matters.

## Priority / Tags

p0, smoke, desktop, feature-name

## Initial State

- Fresh Electron user data directory, unless persistence is the subject.
- Mock CLI scenario requirements.
- Any required workspace, dialog response, policy response, or seeded data.

## Intent Steps

1. User-centered action.
2. Observable action or navigation.
3. Verification step.

## Expected Outcomes

- Concrete visible UI state.
- Relevant persisted/logged behavior when observable through test logs.
- Recovery or closing behavior where appropriate.
- No uncaught renderer error.

## Evidence Required

- Screenshot names or moments worth capturing.
- Console and page error output.
- Any mock CLI log evidence worth checking.

Quality Bar

• Use p0 for release-blocking smoke flows; use p1 for important surfaces; use p2 for edge cases and exploratory coverage.
• Make each expected outcome externally observable from the running app, screenshots, logs, or console output.
• Prefer stable human-facing terms: accessible labels, menu names, panel titles, visible messages.
• Include negative or recovery checks for risky flows: cancellation, empty states, disabled actions, denied approvals, failed config, or restart.
• Keep each plan short enough for an agent to execute in one focused run.

Output

After writing or updating a plan, report:

• The file path.
• The workflow covered.
• Any mock scenario assumptions.
• The exact command to dry-run and execute it:
  • bun run test:intent:e2e -- test-plan/<name>.md --dry-run
  • bun run test:intent:e2e -- test-plan/<name>.md