sp-write-test-plan

name	sp-write-test-plan
description	Use when the user asks for an integration test plan for a multi-component, failure-prone, or historically fragile system.

Integration Test Planning

Write a reasoning-driven test plan in FlickNote. Cover expected behavior and adversarial failure modes.

Explore Reality

• Read the current implementation and existing tests.
• Search prior designs, research, and bug reports with flicknote find <keywords>.
• Map entry points, state transitions, external systems, concurrency, retries, and partial failures.
• Confirm the target repository and exact test surface.

Discuss the proposed test boundary with the user before writing the plan if it is not already agreed.

Constructive Coverage

Define:

• Major happy-path lifecycles
• Boundary and empty inputs
• Defaults and missing configuration
• State invariants and security preconditions
• Idempotency and resource lifecycle rules

Adversarial Coverage

Run three passes:

1. Prior failures: turn each relevant historical bug class into a regression test.
2. Seam walk: concurrency, retries, partial failures, idempotency, timeouts, corruption, edge dates/numbers, hostile inputs.
3. Red team: write concrete failure hypotheses and the test that would falsify each one.

Distinguish:

• Planned: test is needed; no proof of failure
• Confirmed broken: code or evidence proves current behavior is wrong
• Dismissed: investigated and ruled out with a reason

Persist to FlickNote

Save the plan in orientation:

flicknote add --project orientation

Structure:

Test Plan:

Scope and anti-goals

Current implementation

Constructive cases

Prior-failure regressions

Seam walk

Red-team hypotheses

Test environment and fixtures

Execution order

Exit criteria

Risks and unreachable coverage

Each test case must identify the setup, action, expected result, proof source, and cleanup. Record confirmed bugs in the same note; do not create tasks automatically.

Validate

Re-read the note as the test implementer:

• Are fixtures and environments obtainable?
• Can each result be attributed to the layer under test?
• Are negative controls actually achievable?
• Are paths, commands, evidence, and cleanup exact?
• Is unreachable third-party coverage stated plainly?

Update the FlickNote note, then return its ID and a concise summary. Stop; do not execute the plan or invoke another skill.