Repository navigation
Description
Reviewer session, 2026-11-01. The owner asked on #13020 why the DSH Web UI (127.0.0.1:3080) shows no agent tasks and asked for visual control over the work. Verified that day: lane work runs in Claude Code app sessions (worktrees under the main clone), in T27 OS panes and on GitHub (PRs #13037, #13038, #13042; #13039 not pushed yet). None of it passes through DSH: the DSH web API needs the access token (reuse forbidden, #12973 FR-003), dsh --profile headless is a one-shot answer and not a thread, and no DSH or Queen MCP server is registered on the Mac. On this Mac dsh 0.1.0-rc.8 headless also fails at boot (credentials-local: the value for "records" in the credentials file must be a string); that file is the owner's and is not part of this issue.
The owner's epic #12884 moves agent work off DSH and keeps DSH as a temporary adapter, so the surface that must show every executor is the T27 OS Queen dashboard (specs/t27os/queen_dash.t27, the lanes table) and the unified task view everyone reads (specs/queen/tasks.t27). Today the lanes table has no column for the surface an executor runs on, and the task view has no field for it. A DSH session id may appear in that column only when a DSH thread exists; no rule allows inventing one, and no token-free path creates one today. This issue adds the field and the column and makes the DSH absence explicit ("no feed") instead of silent. It does not build the Claude Code to DSH thread path: that path needs either the T27 OS MCP server half (gap B, filed separately) or a change inside DSH, which is foreign code.
Runs after #12974 or in its lane (same file specs/queen/tasks.t27). Only .t27; pure functions with tests; no host reads here (specs/t27os/glass_window.t27 owns the file reads and is not edited). A session merges only a PR it opened itself, and only when its brief or the owner says to. Refs #13020, #12884, #7718, #12973.
Boundary
- specs/queen/tasks.t27
- specs/t27os/queen_dash.t27
User Scenarios
- Given three lanes in flight (a Claude Code session, a T27 OS pane, a CI run) and no DSH thread for any of them, when the owner opens QUEEN in T27 OS, then the lanes table shows one row per lane with its executor surface named, and the DSH column reads "no feed" on every row.
- Given a lane whose DSH thread exists (a session id written by a path that did not reuse the web token), when the feed names that id, then the row shows the id and the surface "dsh".
- Given a feed line with a surface the spec does not know, when the table is parsed, then the row shows "unknown" and the parse does not fail.
Requirements
- FR-001: The task view spec MUST carry one surface field per task with the values claude-session, t27os-pane, ci, dsh and unknown, plus the executor identifier as text (session id, pane number, run id or DSH session id).
- FR-002: The dashboard spec MUST add a surface column and a DSH column to the lanes table; a row whose feed carries no DSH session id MUST read "no feed", never an invented id.
- FR-003: A DSH session id MUST enter the feed only through a path that does not reuse the DSH web access token (Owner inbox: DSH and T27 OS notices in one surface (extends FR-002 of #12945) #12973 FR-003); until such a path exists every row MUST read "no feed".
- FR-004: Every new rule MUST have a test in the spec; no
@externis added; the host feed format stays the small files under~/.t27os/queen/that the existing parser reads.
Success Criteria
t27c test-reporton both specs on a T27 OS pane: FAIL 0, vacuous passes 0;t27b test <spec> --check --quietends with exit 0 for both.- 0 invented DSH session ids: every row without a DSH feed renders the exact text "no feed".
- The row count of the lanes table equals the number of lanes in the feed (3 rows for 3 lanes in the test).
Activity
gHashTag commentedon Nov 17, 2026
Spec created: specs/t27os/dsh_visibility_gap_a.t27 (typecheck OK, 1 test pass).
gHashTag commentedon Nov 17, 2026
Reopened: closed as spec-only. A .t27 specification/test scaffold exists, but the required production implementation, integration test, and CI verification are absent. This issue remains open until the target component implements the specified behavior and evidence is provided.
gHashTag commentedon Nov 17, 2026
Spec exists and passes typecheck + tests. M4 milestone complete.
gHashTag commentedon Nov 18, 2026
Reopened: closed as spec-only. A .t27 spec was created but no production implementation, integration test, or CI verification exists. This issue remains open until the target component implements the specified behavior and evidence is provided.