QA Executor

Use this skill to execute or guide execution of approved QA test cases. Treat execution_result.json as the source of truth, then update the original test case source incrementally, render optional reports only when requested, and produce reporter handoffs or automation signals from canonical state.

Core Rule

Do execution work only. Do not create test plans, invent test cases from raw requirements, submit final issues, write final automation scripts, alter planner approval, invent credentials, invent selectors, invent endpoints, invent DB queries, or invent fixtures. If tools or safe access are unavailable, produce manual_instruction output with structured steps and result templates.

Plane, Notion, spreadsheets, documents, and other source artifacts are synced workflow surfaces, not the source of truth. Keep execution_result.json canonical and render source row/item updates, Plane/Notion comments, status moves, worklogs, links, wiki pages, and optional report pages from that state.

Default to mandatory incremental source updates. Read one test case or a small ordered batch, execute it, update the same source row/item/section when writable, record the update result in execution_result.json, then continue. When a selected test case came from Notion, Excel, Google Sheets, Google Docs, Word, Markdown, JSON, Plane, a qa-planner handoff, a qa-executor handoff, or any other source with a locator, source status update is required. If the source cannot be updated, record a source_update_gap; do not silently skip it. Do not create a new execution report document/page/workbook by default. Create a new report only when the user explicitly asks for report output or the package sets report output policy to enabled. Report creation, handoff contracts, Plane comments, and execution_result.json are additive only; they never replace source test case status updates.

Input Paths

Accept four input paths.