name: opencode-test-runner description: | Testing phase instructions for OpenCode artefact review: generate test cases, narrate dry-runs, iterate on failures. Invoked by opencode-simulated-test-runner agent (preferred) or opencode-reviewer agent (fallback). Not for direct use. metadata: version: "1.0.2" maintainers: [bew]

OpenCode Artefact Tester

Goal

Run structured dry-run testing for a draft OpenCode artefact. Report failures back to the caller.

Steps

1. Generate test cases covering: happy path, common edge case, and failure mode.

Use count from task description if specified, otherwise default to 2–4. Present all test cases to user at once. Ask user to add, replace, or remove any before proceeding.

1. Run every test case in a single pass: narrate all dry-runs in one response.

Never stop between test cases to ask for confirmation.

1. If any test case is wrong: report which test cases failed and why, then return to caller.
2. When all test cases pass: report "All test cases passed." — no other output.

Rules

• Never claim to actually execute anything — simulation only.

Exception: if the artefact under test invokes a real executable to produce a value the test depends on (e.g. a path resolver script), ask the user before running it.

• Never output file contents as confirmation or summary.