Real-app MCP runtime transcript Environment: Linux, Chromium 156, 1280x800, light theme, fresh unseeded development state. Synthetic thread: Pin State Check; Replay (dev), multi-turn; prompt: Pin state check. Driver: MCP initialize, organize(pin), read, list(allProjects), organize(unpin), read, list(allProjects). Base 9ec71438cd43b0f3179ab5ca0c0a2a783fb8280a: ASSERTION_FAILED: expected [{"read":"FIELD ABSENT","list":"FIELD ABSENT"},{"read":"FIELD ABSENT","list":"FIELD ABSENT"}] to equal [{"read":true,"list":true},{"read":false,"list":false}] ASSERTION_FAILED: expected [{"read":"FIELD ABSENT","list":"FIELD ABSENT"},{"read":"FIELD ABSENT","list":"FIELD ABSENT"}] to equal [{"read":true,"list":true},{"read":false,"list":false}] Head fc3ec3ea09c3ecbbe835855773fc849bdef08ecf: Tests 1 passed (1) Observed head responses: pin -> read.pinned=true, list.pinned=true; unpin -> read.pinned=false, list.pinned=false. Each final still matched its preceding screenshot byte-for-byte after a 500ms interval. Response panels are labelled driver instrumentation displaying selected actual MCP response fields; absent fields are not added.