[rigel-mintaka]

This PR is part of a stack containing 4 PRs:

1. main
2. feat(ui): board summary badges in the sidebar and FleetPane (RIG-5177) #2290
3. "feat(board): show asks and their forge mirror state on boards (RIG-5173)" (this PR)
4. feat(ui): add the agent board panel and tasks search over ListAgentBoards (RIG-5409) #2300
5. feat(mirror): agent forge_mirror_messages tool with real-session attribution (RIG-5174) #2319

Stack

• main
• feat(store): status board tables, patch merge and board reads (RIG-5170) #2212 board store (RIG-5170)
• feat(comms): carry an optional issue or PR ref on asks (RIG-5173) #2228 Ask.ref (RIG-5173)
• feat(board): derive board refs, PR gates and stacks from stored rows (RIG-5171) #2231 board derive (RIG-5171)
• feat(mirror): mirror asks and topics to forge issues and PRs (RIG-5181) #2275 ask mirroring (RIG-5181)
• feat(board): agent board RPCs, roster summary and board events (RIG-5172) #2285 board RPCs (RIG-5172)
• feat(ui): board summary badges in the sidebar and FleetPane (RIG-5177) #2290 board badges UI (RIG-5177)
• feat(board): show asks and their forge mirror state on boards (RIG-5173) #2292 this PR, the tip

What

This is the board-asks half of T5. Design: docs/designs/agent/compass-agent-status-record/design.md §T5, §D5, §D6 "Delivery".

• ListAgentBoards fills each board's asks from AsksByAuthor(viewer = caller, …). A board lists every pending ask plus asks answered in the last 24h, pending first, at most 20. An ask in a channel the caller can't see is left out.
• Each ask gets mirror_state and mirror_url. The URL is set only when the mirror is done. Mirror rows for all listed boards are fetched in one batched lookup. When an ask has several rows, the agent's automatic row wins, then the newest row.
• The UpdateAgentBoard response fills the same fields.
• A board write re-drives the agent's open mirror rows through MirrorDriver.DriveAgent (OpenForgeMirrors, then a drive and a recheck schedule per row), in the agent's tenant context.
• Spec: the ListAgentBoards requirement now covers asks, with a scenario for an ask in a channel the caller can't see. The mirroring requirement now covers the re-drive on a board write.

Tests

• Comms pgtests:
  • TestListAgentBoardsIncludesVisibleAsks: viewer clipping and order.
  • TestListAgentBoardsOmitsOldAnsweredAsk
  • TestListAgentBoardsIncludesForgeMirrorStates: done with a URL, failed without one.
  • TestBoardWriteDrivesAgentMirrors: uses a recording driver.
• Server pgtest TestForgeMirrorDriveAgentDeliversOpenAskRow.
• Red check: with the ask fill stubbed out, the visible-asks and mirror-state tests fail.

Review

• Round 1: 1 medium. Per-board mirror queries, now one batched lookup.
• Round 2: 1 medium. The spec said "newest first", but the design (§D5 UI: "Asks, pending first") and T4's query put pending asks first. The spec wording was fixed.
• Both are resolved.

Gates:

• golangci-lint 0 issues (default and pgtest tags);
• nilaway clean;
• unit tests;
• pgtest internal/comms, server;
• rumdl;
• the pre-push moon ci passed 80 of 80.

[linear-code]

RIG-5173

RIG-5172

[github-actions]

Compass engineering docs preview: https://compass-comms-rig-5173-board.compass-eng-docs.pages.dev

Deployed from compass-comms/rig-5173-board-asks at e48fecd.

Changed pages:

• docs/specs/product/compass.md

[rigel-mintaka]

Review loop, head d356d42b8c24

Round	Findings	Status
1	1 medium: per-board mirror queries	fixed: one batched AskForgeMirrors per list or update
2	1 medium: spec said "newest first" while the query and design put pending first	fixed: spec matches design §D5/§D9

All clear. CI is green on this head. The pre-push moon ci passed 80 of 80.

[rigel-mintaka]

Rebased onto #2290 c94a2da after the line moved onto main. New head 1079648; the diff over its base is unchanged (7 files). Focused pgtests pass; pre-push moon ci 80/80.

[rigel-mintaka]

Rebased with the line onto main 309cd3f. New head 1e8efc4, on #2290 c13888b; jj interdiff vs 1079648 is empty. Pre-push moon ci 80/80.

[rigel-mintaka]

Rebased with the line onto main 33db616. New head 75ad4ce on #2290 32ddd1d; jj interdiff vs 1e8efc4 is empty. Pre-push moon ci 80/80.

[rigel-mintaka]

Head note, 1d3ec69a

• Patch unchanged since review. git range-diff d356d42b^..d356d42b bfd71872..1d3ec69a gives 1: d356d42b8 = 1: 1d3ec69a6, and the range-diff against the previous head 75ad4ce8 (on 32ddd1db) is also =. The only differences come from rebasing onto the moved parents (feat(mirror): mirror asks and topics to forge issues and PRs (RIG-5181) #2275 and feat(board): agent board RPCs, roster summary and board events (RIG-5172) #2285 merged to main; feat(ui): board summary badges in the sidebar and FleetPane (RIG-5177) #2290 is now bfd71872).
• Review loop: clean at d356d42b (see the review-loop note above). An identical patch needs no new round.
• CI: 11 pass, 0 fail, 0 pending at 1d3ec69a. GitHub reports MERGEABLE; BLOCKED only on the required review.

[rigel-mintaka]

Rebased with the line onto main 888ac07. New head e48fecd on #2290 cad170c. git range-diff against 1d3ec69 shows one hand edit: go/internal/comms/board_asks_pgtest_test.go now passes t.Context() to NewComms, matching the ctx parameter main added. The rest of the patch is identical. Gates on the rebased line: go vet with pgtest, golangci-lint with and without pgtest (0 issues), nilaway, and proto drift all clean. pgtest: comms, runnerhub, server, store and mirror ok. Pre-push moon ci passed. The remaining review-loop delta is a mechanical one-argument change, so no new round.

[rigel-mintaka]

CI is green at e48fecd: 11 pass, 0 fail, 0 pending. Attempt 2 failed in pgtest setup on an external cache.nixos.org substitution error, recorded on RIG-5463; attempt 3 passed. The review round on the rebase delta (1d3ec69 → e48fecd, where NewComms now takes ctx) ran 4 lenses with 0 findings.