diff --git a/.pi/memory-vault/packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in.md b/.pi/memory-vault/packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in.md
new file mode 100644
index 0000000..918e950
--- /dev/null
+++ b/.pi/memory-vault/packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in.md
@@ -0,0 +1,78 @@
+---
+type: context
+id: context.paperclip-harness.both-defects-fixed-and-the-settings-editor-is-in
+title: both-defects-fixed-and-the-settings-editor-is-in
+status: active
+source_of_truth: false
+freshness: current
+last_reviewed: 2026-11-10
+tags:
+  - pack/paperclip-harness
+  - agent-memory/context
+  - autolearn
+  - rich-discovery
+  - repo/both-defects-fixed-and-the-settings-editor-is-in
+---
+
+# both-defects-fixed-and-the-settings-editor-is-in
+
+## Rich discovery update (2026-11-10 06:23:51)
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Source paths observed
+
+- `omni/deepseek/deepseek-v4-flash`
+
+## Evidence
+
+- Confidence: 84%
+- Reason: Update component/repository context with detailed discovered structure and configuration
+
+## Related
+
+- [[packs/paperclip-harness/00-system/pi-agent/memory-manifest|Memory Manifest]]
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status.md b/.pi/memory-vault/packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status.md
new file mode 100644
index 0000000..31fce45
--- /dev/null
+++ b/.pi/memory-vault/packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status.md
@@ -0,0 +1,85 @@
+---
+type: context
+id: context.paperclip-harness.paperclip-tui-workspaces-panels-and-tools-status
+title: Paperclip TUI workspaces/panels and tools status
+status: active
+source_of_truth: false
+freshness: current
+last_reviewed: 2026-11-10
+tags:
+  - pack/paperclip-harness
+  - agent-memory/context
+  - paperclip
+  - tui
+  - workspaces
+  - panels
+  - docs
+  - ocr
+---
+
+# Paperclip TUI workspaces/panels and tools status
+
+Paperclip TUI organizes UI into workspaces containing panels. As of commit `ec7e907` (`docs/WORKSPACES.md`): **22 panels built across 8 workspaces**, with **one panel missing** and **six tool panes** still to wire.
+
+Remaining work:
+1. **Work>review** — the ocr verdict shown beside the diff. This is the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf, or ocr in the current workspace; `esc` closes it. The six tools: Lazygit, Hunk, Yazi, Neovim, fzf, ocr.
+
+`docs/WORKSPACES.md` is the tracking doc for panel/tool-workspace status and should be updated as panels are built.
+
+## Detailed context from completed turn
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Evidence
+
+- Confidence: 80%
+- Reason: Architecture/status details from final summary and commit reference; useful for future work on remaining panels.; enriched with final-answer detail because persistence should be reusable
+
+## Related
+
+- [[packs/paperclip-harness/00-system/pi-agent/memory-manifest|Memory Manifest]]
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/20-context/tmp.md b/.pi/memory-vault/packs/paperclip-harness/20-context/tmp.md
index 9ff6a54..2cb6c17 100644
--- a/.pi/memory-vault/packs/paperclip-harness/20-context/tmp.md
+++ b/.pi/memory-vault/packs/paperclip-harness/20-context/tmp.md
@@ -466,3 +466,33 @@ Say which of the three you want next, and I will build only that.
 - session: [[packs/paperclip-harness/80-sessions/rich-discovery-the-full-list-at-the-end-of-the-project-with-|Rich discovery: the-full-list-at-the-end-of-the-project-with-what-exists-tod 2026-11-10 05:59:07]]
 - context: [[packs/paperclip-harness/20-context/tmp|the-full-list-at-the-end-of-the-project-with-what-exists-tod]]
 - observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|the-full-list-at-the-end-of-the-project-with-what-exists-tod repository structure and configuration patterns]]
+
+
+---
+
+## Update (2026-11-10 06:23:51)
+
+The Paperclip terminal UI (TUI) has an "Agents" pane with four tabs: roster, runs, log, and settings. It tracks agents (researcher, System, Hub, Chief of Staff, Builder, Documenter), their states (idle/paused), spend vs. budget, and run history.
+
+Key behaviors:
+- **runs tab**: project-scoped to the current project's tasks by default (resolved through each run's task reference, shown as e.g. `AUD-166`). `A` toggles between "this project" and company-wide runs. Company-wide total at time of work: 281 runs.
+- **log tab**: renders run logs. Logs are a stream of token deltas; the viewer merges deltas into paragraphs while tool calls, prompts, and turn markers each become single lines (e.g. `toolcall_start`, `toolcall_end`, `assistant:` lines).
+- **settings tab**: editable fields for each agent — role, job title, model, monthly budget, heartbeat. `enter` edits a field inline, `esc` cancels, heartbeat is a toggle. Saves go through one `PATCH /api/agents/{id}` that merges into the agent's existing config (does not wipe working folder or instructions). `p`/`P`/`w` pause/resume/wake the selected agent.
+- **Navigation**: each tab has its own cursor; `enter` opens detail (e.g. run log). Roster is sorted by name so cursor stays on the same agent after reload; `▸` marks the selected row.
+
+The pause control writes `pausedAt` and `pauseReason` (e.g. `manual` from a board key) to the agent status.
+
+## Evidence
+
+- Confidence: 85%
+- Reason: Evidence-backed overview of the Agents pane architecture and behaviors from the turn's tool output and final summary.
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups.md b/.pi/memory-vault/packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups.md
index 0a3093e..0fc7597 100644
--- a/.pi/memory-vault/packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups.md
+++ b/.pi/memory-vault/packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups.md
@@ -280,3 +280,35 @@ Follow-up: user asked to pick item 1..5; only build the chosen item.
 - session: [[packs/paperclip-harness/80-sessions/paperclip-harness-reverted-pane-split-interface-gap-discover|Paperclip-harness — reverted pane split, interface gap discovery, and PLAN-INTERFACE roadmap]]
 - session: [[packs/paperclip-harness/80-sessions/rich-discovery-home-2026-11-10-05-41-48|Rich discovery: home 2026-11-10 05:41:48]]
 - context: [[packs/paperclip-harness/20-context/home|home]]
+
+
+---
+
+## Update (2026-11-10 06:23:51)
+
+Delivered and committed (`ec7e907`: "Agents pane: project-scoped runs with selection, roster cursor, editable settings").
+
+Changes:
+- Runs tab now project-scoped by default with task column, `A` toggles company-wide.
+- Added per-tab cursors for runs and roster; `enter` opens the selected run's log; selection highlighted, scroll-follows-cursor.
+- Log viewer now merges token deltas into paragraphs; tool calls/prompts/turn markers as single lines.
+- New settings tab (4th tab): edit role, job title, model, monthly budget, heartbeat via inline edit (`enter` saves, `esc` cancels, heartbeat toggle). Saves are single `PATCH /api/agents/{id}` merging into existing config; verified config preserved after save.
+- `p`/`P`/`w` pause/resume/wake act on the selected agent.
+- Updated `docs/WORKSPACES.md` (22 panels across 8 workspaces, 1 missing panel, 6 tool panes to wire).
+
+Validation: build passed (`build=0`); manual tests of run selection/open, settings edit/save, and pause/resume round-trip.
+
+## Evidence
+
+- Confidence: 90%
+- Reason: Completed, verified deliverable with commit hash, artifacts, and validation.
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns.md b/.pi/memory-vault/packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns.md
index f3d446c..fd0e82d 100644
--- a/.pi/memory-vault/packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns.md
+++ b/.pi/memory-vault/packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns.md
@@ -979,3 +979,172 @@ Say which of the three you want next, and I will build only that.
 - session: [[packs/paperclip-harness/80-sessions/rich-discovery-the-full-list-at-the-end-of-the-project-with-|Rich discovery: the-full-list-at-the-end-of-the-project-with-what-exists-tod 2026-11-10 05:59:07]]
 - context: [[packs/paperclip-harness/20-context/tmp|the-full-list-at-the-end-of-the-project-with-what-exists-tod]]
 - observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|the-full-list-at-the-end-of-the-project-with-what-exists-tod repository structure and configuration patterns]]
+
+
+---
+
+## Update (2026-11-10 06:23:51)
+
+Three defects were diagnosed in the Paperclip TUI Agents pane:
+- **Runs list was company-wide** (281 runs) instead of project-scoped; fixed by resolving each run's task reference to the current project and showing the task ID (e.g. `AUD-166`). Added `A` to toggle project/all scope.
+- **Scrolling instead of selecting** (runs tab): the list had no cursor, so arrow keys only moved the viewport. Fix: per-tab cursor, highlighted selection row, scroll only to keep cursor visible. `enter` opens the selected run's log.
+- **Roster could not be walked**: same root cause (no cursor); fixed with per-tab cursor and name-sorted roster so selection is stable across reloads.
+
+Also fixed log readability: raw token-delta stream was gathered into paragraphs while tool calls/prompts/turn markers became single lines.
+
+## Detailed observations from completed turn
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Evidence
+
+- Confidence: 85%
+- Reason: Diagnosis and root-cause findings from tool output and summary; valuable for future Agents pane work.; enriched with final-answer detail because persistence should be reusable
+
+
+---
+
+## Update (2026-11-10 06:23:51)
+
+The Builder agent was observed paused with `status: paused pausedAt: 2026-11-10T05:56:47.518Z reason: manual`, caused by a keystroke test on the board key (not the user). Resuming via the pause/resume control returned HTTP 200 and the agent returned to `status: idle`.
+
+Caveat for future agents: pause/resume/wake keys (`p`/`P`/`w`) act on the currently selected agent and can unintentionally pause a live agent; always confirm the selection before keying. Pause state persists in agent status and can be resumed via the same control.
+
+## Detailed observations from completed turn
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Evidence
+
+- Confidence: 80%
+- Reason: Concrete observed state, API behavior, and operational caveat.; enriched with final-answer detail because persistence should be reusable
+
+
+---
+
+## Update (2026-11-10 06:23:51)
+
+## Structural/configuration observations (2026-11-10 06:23:51)
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Source paths observed
+
+- `omni/deepseek/deepseek-v4-flash`
+
+## Evidence
+
+- Confidence: 82%
+- Reason: Captured factual repo structure/configuration patterns from rich discovery
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui.md b/.pi/memory-vault/packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui.md
new file mode 100644
index 0000000..cfcd5b6
--- /dev/null
+++ b/.pi/memory-vault/packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui.md
@@ -0,0 +1,90 @@
+---
+type: runbook
+id: runbook.paperclip-harness.editing-agent-settings-and-pausing-resuming-in-tui
+title: Editing agent settings and pausing/resuming in TUI
+status: active
+source_of_truth: false
+freshness: current
+last_reviewed: 2026-11-10
+tags:
+  - pack/paperclip-harness
+  - agent-memory/runbook
+  - paperclip
+  - tui
+  - runbook
+  - agents
+  - settings
+  - pause
+---
+
+# Editing agent settings and pausing/resuming in TUI
+
+Trigger: need to change an agent's role/job title/model/budget/heartbeat, or pause/resume/wake an agent via the Paperclip TUI.
+
+Prerequisites: TUI running, Agents pane available.
+
+Procedure:
+1. Navigate to Agents>settings tab.
+2. Move cursor to the target agent on the roster; confirm `▸` is on the intended row (pause keys act on the selected agent — misselection can pause a live agent).
+3. `enter` on a field to edit inline; type value; `enter` saves, `esc` cancels. Heartbeat is a toggle.
+4. Save issues a single `PATCH /api/agents/{id}` that merges into existing config (does not wipe working folder/instructions). Model change applies to the agent's next run, not an in-flight run.
+5. To pause/resume/wake the selected agent press `p` / `P` / `w` respectively.
+
+Verification / rollback: read the agent record back after saving to confirm fields persisted. If you accidentally pause an agent, resume with `P`; expect resume to return HTTP 200 and status to return to idle.
+
+## Detailed procedure from completed turn
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Evidence
+
+- Confidence: 80%
+- Reason: Repeatable procedure with a safety caveat learned during the turn.; enriched with final-answer detail because persistence should be reusable
+
+## Related
+
+- [[packs/paperclip-harness/00-system/pi-agent/memory-manifest|Memory Manifest]]
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/.pi/memory-vault/packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is.md b/.pi/memory-vault/packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is.md
new file mode 100644
index 0000000..c19a9bf
--- /dev/null
+++ b/.pi/memory-vault/packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is.md
@@ -0,0 +1,81 @@
+---
+type: session
+id: session.paperclip-harness.rich-discovery-both-defects-fixed-and-the-settings-editor-is
+title: Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51
+status: active
+source_of_truth: false
+freshness: current
+last_reviewed: 2026-11-10
+tags:
+  - pack/paperclip-harness
+  - agent-memory/session
+  - autolearn
+  - rich-discovery
+  - repo/both-defects-fixed-and-the-settings-editor-is-in
+---
+
+# Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51
+
+Sanitized rich discovery snapshot captured automatically from the completed turn.
+
+
+## Source paths observed
+
+- `omni/deepseek/deepseek-v4-flash`
+
+## Final answer snapshot
+
+Both defects fixed, and the settings editor is in.
+
+## The three things you reported
+
+**Runs were every run in Paperclip.** Correct — 281 company-wide. They are now scoped to this project's tasks, resolved through each run's own task reference, and the task column shows `AUD-166` instead of a bare identifier. `A` toggles between "this project" and "everything in the company", because both are legitimate questions but not the same one.
+
+**Scrolling instead of selecting.** The run list had no cursor at all — arrows only moved the viewport. There is now one cursor per tab, the selected row is highlighted, and the list scrolls only to keep the cursor visible. `enter` opens that run's log. Tested: picked the second run, its log opened.
+
+**Roster could not be walked.** Same cause, same fix. The roster is sorted by name so the cursor stays on the same agent after a reload, and `▸` marks the row.
+
+The log itself also became readable: it is a stream of token deltas, so deltas are gathered back into paragraphs while tool calls, prompts and turn markers become single lines:
+
+```
+15:55:57 toolcall_start
+15:55:57 {"command": "pwd && ls -la && git log --oneline -10 2>/dev/null"}
+15:55:58 toolcall_end
+15:55:58 assistant: I'll start by inspecting the workspace…
+```
+
+## Agents>settings is built
+
+Fourth tab. Role, job title, model, monthly budget, heartbeat. `enter` edits the chosen field inline, `esc` cancels; heartbeat is a toggle. Every save is one `PATCH /api/agents/{id}` that merges into that agent's existing configuration, so a model change does not wipe its working folder or instructions — verified by reading the record back after a save.
+
+`p`, `P`, `w` pause, resume and wake, acting on the agent chosen on that tab.
+
+**One thing you should know:** `Builder` was paused at 16:56 with `pauseReason: manual`, from the board key — that was my keystroke test, not you. I have resumed it and it is idle again. Sorry for the scare.
+
+## Still missing, in order
+
+1. **Work>review** — the ocr verdict beside the diff. That is now the only missing panel.
+2. **Tool panes** — `t` opens Lazygit, Hunk, Yazi, Neovim, fzf or ocr in the current workspace; `esc` closes it.
+
+`docs/WORKSPACES.md` is updated: 22 panels built across 8 workspaces, one panel missing, six tool panes to wire. Committed as `ec7e907`.
+
+Which of those two next?
+
+## Evidence
+
+- Confidence: 86%
+- Reason: Preserve full sanitized discovery detail from a rich investigation turn
+
+## Related
+
+- [[packs/paperclip-harness/00-system/pi-agent/memory-manifest|Memory Manifest]]
+
+## Related learned memories <!-- same-turn-2026-11-10 06:23:51 -->
+
+- context: [[packs/paperclip-harness/20-context/tmp|Paperclip TUI Agents pane]]
+- context: [[packs/paperclip-harness/20-context/paperclip-tui-workspaces-panels-and-tools-status|Paperclip TUI workspaces/panels and tools status]]
+- observation: [[packs/paperclip-harness/60-observations/tmp-repository-structure-and-configuration-patterns|Agents pane defect causes: missing cursors, unscoped runs]]
+- runbook: [[packs/paperclip-harness/70-runbooks/editing-agent-settings-and-pausing-resuming-in-tui|Editing agent settings and pausing/resuming in TUI]]
+- action: [[packs/paperclip-harness/40-actions/2026-11-10-wrote-docs-ui-layout-md-and-three-live-layout-mockups|Agents pane: scoped runs, cursors, settings editor]]
+- session: [[packs/paperclip-harness/80-sessions/rich-discovery-both-defects-fixed-and-the-settings-editor-is|Rich discovery: both-defects-fixed-and-the-settings-editor-is-in 2026-11-10 06:23:51]]
+- context: [[packs/paperclip-harness/20-context/both-defects-fixed-and-the-settings-editor-is-in|both-defects-fixed-and-the-settings-editor-is-in]]
diff --git a/README.md b/README.md
index 12fadd6..efe5d3d 100644
--- a/README.md
+++ b/README.md
@@ -59,16 +59,24 @@ program running in the Board workspace.
 | Workspace>pane | shows | keys |
 |---|---|---|
 | Shell>shell | a prompt in the project folder | — |
-| Board>board | tasks by group · the chosen task · needs you | ↑↓ j k · enter · a · u · / · tab |
-| Work>work | runs · change · facts | [ ] tab · ↑↓ j k · g G |
+| Board>board | tasks by group · the chosen task · needs you | ↑↓ j k · enter · a · u · / · t (tools) · tab |
+| Work>work | runs · change · facts · review | [ ] tab · ↑↓ j k · g G · t (tools) |
 | Reply>reply | the task · the thread · the reply box | enter · s · esc |
 | New task>new | the form · what a task needs · what will be sent | tab · enter · ctrl+n |
 | Project>project | facts · costs · settings | [ ] tab · ↑↓ j k |
 | Repair>repair | health · layout · rebuild log | [ ] tab · ↑↓ j k · R · B |
+| Agents>agents | roster · runs · log · settings | [ ] tab · ↑↓ j k · enter · p P w · A |
 
 Choosing a task on the board and pressing Enter writes it into the project's
 `.harness/harness.toml`, and every other workspace follows within a second.
 
+### Tool panes
+
+`t` in `Board>board` or `Work>work` opens a chooser: shell, Lazygit, Hunk, Yazi, Neovim,
+fzf, ocr. The tool opens as a pane beside the one you are in, in the same workspace, and
+`esc` closes the chooser. Quitting the tool (its own key) closes its pane. Nothing is added
+to the layout, so the eight workspaces stay as they are.
+
 ### Refresh and rebuild
 
 - **`r`** — read the data again in this panel. Nothing closes, nothing moves, the
diff --git a/bin/harness b/bin/harness
index be723d5..76b5ec8 100755
Binary files a/bin/harness and b/bin/harness differ
diff --git a/docs/WORKSPACES.md b/docs/WORKSPACES.md
index e605142..dd21661 100644
--- a/docs/WORKSPACES.md
+++ b/docs/WORKSPACES.md
@@ -27,7 +27,7 @@ written yet.
 | `Work>runs` | run timeline and activity, with cost per run | built |
 | `Work>change` | worktree, branch, base, commits, changed files, the diff | built |
 | `Work>facts` | task, state, worktree path, branch, base, cost | built |
-| `Work>review` | the last ocr verdict for this task: findings, blocking ones in red, and whether a merge would be refused | **missing** |
+| `Work>review` | the verdicts recorded on this task: review findings, refused merges, merged results, blocking ones in red | built |
 
 ## 4. Reply
 
@@ -78,19 +78,20 @@ Yazi, fzf, Neovim, Lazygit, Hunk and anything else fits without being rebuilt by
 
 | Tool | What it is for | Status |
 |---|---|---|
-| Lazygit | branches, commits, staging, history | installed, **not wired** |
-| Hunk | reading a change as a review | installed, **not wired** |
-| Yazi | files | installed, **not wired** |
-| Neovim | editing a file the task touched | installed, **not wired** |
-| fzf | picking anything | installed, **not wired** |
+| Lazygit | branches, commits, staging, history | wired: `t` in Board>board or Work>work |
+| Hunk | reading a change as a review | wired the same way |
+| Yazi | files | wired the same way |
+| Neovim | editing a file the task touched | wired the same way |
+| fzf | picking anything | wired the same way |
 | shell | any command the interface does not cover | built (`Shell>shell`) |
 | ocr | the review of the current change | installed, behind `bin/review` |
 
 ## Count
 
-- **22 panels built**, across 8 workspaces.
-- **1 panel missing**: `Work>review`.
-- **6 tool panes** to wire: Lazygit, Hunk, Yazi, Neovim, fzf, ocr.
+- **23 panels built**, across 8 workspaces. Nothing is missing from the list above.
+- **tool panes wired** in Board>board and Work>work: `t` opens the chooser, `enter` opens
+  the tool beside the pane, `esc` closes it, and the tool's own quit key closes its pane.
+  The other workspaces can get the same key whenever it is wanted.
 
 Nothing in the missing list changes the layout: each is one more panel inside a pane that
 already exists, or one pane-opener for a tool.
diff --git a/tui/agents.go b/tui/agents.go
index c5bb5b9..58843e3 100644
--- a/tui/agents.go
+++ b/tui/agents.go
@@ -84,7 +84,7 @@ func newAgentsModel(c *paperclip.Client, folder string) agentsModel {
 }
 
 func (m agentsModel) Init() tea.Cmd {
-	return tea.Batch(m.load(), tickEvery(5*time.Second))
+	return tea.Batch(m.load(), tickEvery(5*time.Second), tea.WindowSize())
 }
 
 func (m agentsModel) load() tea.Cmd {
diff --git a/tui/board.go b/tui/board.go
index b42c8cb..050e5c2 100644
--- a/tui/board.go
+++ b/tui/board.go
@@ -49,6 +49,8 @@ type boardModel struct {
 	activity []paperclip.Activity
 	needsTab int
 
+	tools    bool
+	toolAt   int
 	zones    *zone.Manager
 	filter   string
 	typing   bool
@@ -71,7 +73,7 @@ func newBoardModel(c *paperclip.Client, proj paperclip.Project, folder string) b
 }
 
 func (m boardModel) Init() tea.Cmd {
-	return tea.Batch(m.reload(), tickEvery(5*time.Second))
+	return tea.Batch(m.reload(), tickEvery(5*time.Second), tea.WindowSize())
 }
 
 func (m boardModel) reload() tea.Cmd {
@@ -157,6 +159,27 @@ func (m boardModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		return m, nil
 
 	case tea.KeyMsg:
+		if m.tools {
+			switch msg.String() {
+			case "esc", "q", "t":
+				m.tools = false
+				return m, nil
+			case "up", "k":
+				m.toolAt = max(0, m.toolAt-1)
+				return m, nil
+			case "down", "j":
+				m.toolAt = min(len(tools)-1, m.toolAt+1)
+				return m, nil
+			case "enter":
+				t := tools[m.toolAt]
+				m.tools = false
+				if err := openTool(m.folder, t.name); err != nil {
+					m.err = err
+				}
+				return m, nil
+			}
+			return m, nil
+		}
 		if m.showHelp {
 			m.showHelp = false
 			return m, nil
@@ -192,6 +215,9 @@ func (m boardModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		case "?":
 			m.showHelp = true
 			return m, nil
+		case "t":
+			m.tools = true
+			return m, nil
 		case "/":
 			m.typing = true
 			return m, nil
@@ -424,6 +450,7 @@ func (m boardModel) footerLine() string {
 			keyHint("enter", "show it everywhere"),
 			keyHint("a", "assign"),
 			keyHint("u", "unblock"),
+			keyHint("t", "tools"),
 			keyHint("/", "filter"),
 			keyHint("?", "keys"),
 			keyHint("tab", "next panel"),
diff --git a/tui/main.go b/tui/main.go
index 5d72f7e..c35a64d 100644
--- a/tui/main.go
+++ b/tui/main.go
@@ -80,11 +80,9 @@ func main() {
 		if dir == "" {
 			dir, _ = os.Getwd()
 		}
-		issue := *issue
-		if issue == "" {
-			issue = watchedTask(dir)
-		}
-		m = newWorkModel(client, dir, issue)
+		// Pass the flag, not a resolved task: an empty flag means "follow the
+		// task chosen on the Board", and resolving it here would pin the pane.
+		m = newWorkModel(client, dir, *issue)
 
 	case "tree":
 		proj, err := client.ResolveProject(*project)
diff --git a/tui/project.go b/tui/project.go
index 9696860..aca3cec 100644
--- a/tui/project.go
+++ b/tui/project.go
@@ -60,7 +60,7 @@ func newProjectModel(c *paperclip.Client, folder string) projectModel {
 }
 
 func (m projectModel) Init() tea.Cmd {
-	return tea.Batch(m.load(), tickEvery(10*time.Second))
+	return tea.Batch(m.load(), tickEvery(10*time.Second), tea.WindowSize())
 }
 
 func (m projectModel) load() tea.Cmd {
diff --git a/tui/repair.go b/tui/repair.go
index 4f9188e..f128a9e 100644
--- a/tui/repair.go
+++ b/tui/repair.go
@@ -63,7 +63,7 @@ func newRepairModel(c *paperclip.Client, folder string) repairModel {
 }
 
 func (m repairModel) Init() tea.Cmd {
-	return tea.Batch(m.load(), tickEvery(8*time.Second))
+	return tea.Batch(m.load(), tickEvery(8*time.Second), tea.WindowSize())
 }
 
 func (m repairModel) load() tea.Cmd {
@@ -194,7 +194,7 @@ func (m repairModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		return m, nil
 
 	case tickMsg:
-		return m, tea.Batch(m.load(), tickEvery(8*time.Second))
+		return m, tea.Batch(m.load(), tickEvery(8*time.Second), tea.WindowSize())
 
 	case tea.KeyMsg:
 		if m.showHelp {
diff --git a/tui/reply.go b/tui/reply.go
index b853242..ce54553 100644
--- a/tui/reply.go
+++ b/tui/reply.go
@@ -55,7 +55,7 @@ func newReplyModel(c *paperclip.Client, folder, issueID string) replyModel {
 }
 
 func (m replyModel) Init() tea.Cmd {
-	return tea.Batch(m.load(), tickEvery(5*time.Second))
+	return tea.Batch(m.load(), tickEvery(5*time.Second), tea.WindowSize())
 }
 
 func (m replyModel) load() tea.Cmd {
diff --git a/tui/toolpane.go b/tui/toolpane.go
new file mode 100644
index 0000000..41eaee7
--- /dev/null
+++ b/tui/toolpane.go
@@ -0,0 +1,180 @@
+package main
+
+// Tool panes: any terminal program this interface does not cover, opened beside
+// what you were reading and closed when you are done with it.
+//
+// Nothing is rebuilt. Lazygit, Hunk, Yazi, Neovim, fzf and ocr are the tools
+// people already use, and a pane is the correct place for them: they keep their
+// own keys, their own colours and their own way of working.
+//
+// A tool pane is not part of the layout. It is created on demand in the
+// workspace you are in and it disappears when the program exits, so the eight
+// workspaces stay as they are.
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strconv"
+	"syscall"
+)
+
+type tool struct {
+	name    string
+	command string
+	about   string
+}
+
+// tools are offered in this order: the shell first because it can reach
+// anything, then the three that read code, then the review.
+var tools = []tool{
+	{"shell", "exec zsh -i", "a prompt in the project folder"},
+	{"lazygit", "lazygit", "branches, commits, staging, history"},
+	{"hunk", "hunk", "a change read as a review"},
+	{"yazi", "yazi", "files"},
+	{"nvim", "nvim", "editing one file"},
+	{"fzf", "fzf", "picking anything"},
+	{"ocr", "ocr review", "the review of the current change, in the terminal"},
+}
+
+// currentWorkspace finds the workspace this program is drawn in. TUIOS knows;
+// the environment names the pane, and the pane's window says which workspace.
+func currentWorkspace() (string, string, error) {
+	session := os.Getenv("TUIOS_SESSION")
+	pane := os.Getenv("TUIOS_PANE_ID")
+	if session == "" || pane == "" {
+		return "", "", fmt.Errorf("this is not running inside a TUIOS pane")
+	}
+	out, err := exec.Command("tuios", "list-windows", "-s", session, "--json").Output()
+	if err != nil {
+		return "", "", fmt.Errorf("cannot ask TUIOS where this pane is: %w", err)
+	}
+	ws := workspaceOf(string(out), pane)
+	if ws == "" {
+		return "", "", fmt.Errorf("this pane is not in the layout TUIOS reports")
+	}
+	return session, ws, nil
+}
+
+// workspaceOf finds the workspace number of one window in the JSON TUIOS
+// prints. It decodes properly rather than scanning text: a pane id appears in
+// more than one place, and the first match is not always the window.
+func workspaceOf(raw, paneID string) string {
+	var doc struct {
+		Windows []struct {
+			WindowID  string `json:"window_id"`
+			Workspace int    `json:"workspace"`
+		} `json:"windows"`
+	}
+	if err := json.Unmarshal([]byte(raw), &doc); err != nil {
+		return ""
+	}
+	for _, w := range doc.Windows {
+		if w.WindowID == paneID {
+			return strconv.Itoa(w.Workspace)
+		}
+	}
+	return ""
+}
+
+// openTool starts a program as a new pane in the current workspace.
+func openTool(folder, name string) error {
+	var chosen *tool
+	for i := range tools {
+		if tools[i].name == name {
+			chosen = &tools[i]
+		}
+	}
+	if chosen == nil {
+		return fmt.Errorf("no tool called %q", name)
+	}
+	session, ws, err := currentWorkspace()
+	if err != nil {
+		return err
+	}
+	log, _ := os.Create(filepath.Join(os.TempDir(), "paperclip-harness-tool.log"))
+
+	cmd := exec.Command("tuios", "new-window", chosen.name, "-s", session,
+		"--workspace", ws, "--cwd", folder, "--", "/bin/sh", "-c", chosen.command)
+	cmd.Stdout = log
+	cmd.Stderr = log
+	cmd.SysProcAttr = &syscall.SysProcAttr{Setsid: true}
+	if err := cmd.Start(); err != nil {
+		return fmt.Errorf("cannot open %s: %w", chosen.name, err)
+	}
+	go func() {
+		cmd.Wait()
+		if log != nil {
+			log.Close()
+		}
+	}()
+	return nil
+}
+
+// --- the smallest JSON reader that does this one job -------------------------
+
+type scanner struct {
+	s   string
+	pos int
+}
+
+func newScanner(s string) *scanner { return &scanner{s: s} }
+
+// nextObject returns the next {...} block, or "" at the end.
+func (sc *scanner) nextObject() string {
+	for sc.pos < len(sc.s) && sc.s[sc.pos] != '{' {
+		sc.pos++
+	}
+	if sc.pos >= len(sc.s) {
+		return ""
+	}
+	depth := 0
+	start := sc.pos
+	for sc.pos < len(sc.s) {
+		switch sc.s[sc.pos] {
+		case '{':
+			depth++
+		case '}':
+			depth--
+			if depth == 0 {
+				sc.pos++
+				return sc.s[start:sc.pos]
+			}
+		}
+		sc.pos++
+	}
+	return ""
+}
+
+func contains(s, sub string) bool {
+	for i := 0; i+len(sub) <= len(s); i++ {
+		if s[i:i+len(sub)] == sub {
+			return true
+		}
+	}
+	return false
+}
+
+// fieldNumber reads "name": 12 out of an object.
+func fieldNumber(obj, name string) string {
+	key := `"` + name + `":`
+	for i := 0; i+len(key) <= len(obj); i++ {
+		if obj[i:i+len(key)] != key {
+			continue
+		}
+		j := i + len(key)
+		for j < len(obj) && (obj[j] == ' ' || obj[j] == '"') {
+			j++
+		}
+		start := j
+		for j < len(obj) && obj[j] >= '0' && obj[j] <= '9' {
+			j++
+		}
+		if _, err := strconv.Atoi(obj[start:j]); err == nil {
+			return obj[start:j]
+		}
+	}
+	return ""
+}
diff --git a/tui/work.go b/tui/work.go
index a6398b1..d0e95de 100644
--- a/tui/work.go
+++ b/tui/work.go
@@ -44,20 +44,22 @@ type workModel struct {
 
 	tab    int
 	scroll int
+	tools  bool
+	toolAt int
 	readAt time.Time
 
 	showHelp      bool
 	width, height int
 }
 
-var workTabs = []string{"runs", "change", "facts"}
+var workTabs = []string{"runs", "change", "facts", "review"}
 
 func newWorkModel(c *paperclip.Client, folder, issueID string) workModel {
 	return workModel{client: c, folder: folder, issue: issueID, explicit: issueID != ""}
 }
 
 func (m workModel) Init() tea.Cmd {
-	return tea.Batch(m.load(), tickEvery(5*time.Second))
+	return tea.Batch(m.load(), tickEvery(5*time.Second), tea.WindowSize())
 }
 
 func (m workModel) load() tea.Cmd {
@@ -120,14 +122,38 @@ func (m workModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 		return m, nil
 
 	case tickMsg:
-		return m, tea.Batch(m.load(), tickEvery(5*time.Second))
+		return m, tea.Batch(m.load(), tickEvery(5*time.Second), tea.WindowSize())
 
 	case tea.KeyMsg:
+		if m.tools {
+			switch msg.String() {
+			case "esc", "q", "t":
+				m.tools = false
+				return m, nil
+			case "up", "k":
+				m.toolAt = max(0, m.toolAt-1)
+				return m, nil
+			case "down", "j":
+				m.toolAt = min(len(tools)-1, m.toolAt+1)
+				return m, nil
+			case "enter":
+				t := tools[m.toolAt]
+				m.tools = false
+				if err := openTool(m.folder, t.name); err != nil {
+					m.err = err
+				}
+				return m, nil
+			}
+			return m, nil
+		}
 		if m.showHelp {
 			m.showHelp = false
 			return m, nil
 		}
 		switch msg.String() {
+		case "t":
+			m.tools = true
+			return m, nil
 		case "?":
 			m.showHelp = true
 			return m, nil
@@ -225,10 +251,24 @@ func (m workModel) View() string {
 		body = m.runsLines(m.width - 4)
 	case 1:
 		body = m.diffLines(m.width - 4)
+	case 3:
+		body = m.reviewLines(m.width - 4)
 	default:
 		body = m.factsLines(m.width - 4)
 	}
 
+	if m.tools {
+		entries := make([][2]string, 0, len(tools))
+		for i, t := range tools {
+			mark := "  "
+			if i == m.toolAt {
+				mark = "▸ "
+			}
+			entries = append(entries, [2]string{mark + t.name, t.about})
+		}
+		return helpOverlay("open a tool pane here", entries, m.width, m.height)
+	}
+
 	p := panel{
 		title: "the work", width: m.width, height: m.height - 2,
 		tabs: workTabs, active: m.tab, focused: true,
@@ -249,6 +289,7 @@ func (m workModel) View() string {
 	footer := hintLine([]string{
 		keyHint("[ ]", "switch tab"),
 		keyHint("↑↓ or j k", "scroll"),
+		keyHint("t", "open a tool pane"),
 		keyHint("r", "read again"),
 		keyHint("R", "rebuild every pane"),
 		keyHint("q", "leave"),
@@ -355,6 +396,45 @@ func (m workModel) diffLines(width int) []string {
 	return out
 }
 
+// reviewLines shows the verdicts already recorded on the task. A review is
+// written as a comment, so the record lives with the work and survives us.
+func (m workModel) reviewLines(width int) []string {
+	out := []string{"", dimStyle().Render(" verdicts recorded on this task")}
+	if m.task.Identifier == "" {
+		return append(out, mutedStyle2().Render(" no task chosen"))
+	}
+	cs, err := m.client.Comments(m.task.ID)
+	if err != nil {
+		return append(out, lipgloss.NewStyle().Foreground(lipgloss.Color(active.Error)).Render(" "+err.Error()))
+	}
+	found := 0
+	for i := len(cs) - 1; i >= 0 && found < 3; i-- {
+		body := cs[i].Body
+		if !strings.HasPrefix(body, "## Review") &&
+			!strings.HasPrefix(body, "## Merge refused") &&
+			!strings.HasPrefix(body, "## Merged") &&
+			!strings.HasPrefix(body, "## Merge undone") {
+			continue
+		}
+		found++
+		blocking := strings.Contains(body, "critical") || strings.Contains(body, "**high/")
+		colour := active.Success
+		if strings.HasPrefix(body, "## Merge refused") || blocking {
+			colour = active.Error
+		}
+		for _, ln := range wrap(strings.ReplaceAll(body, "\n", " · "), width-2) {
+			out = append(out, " "+lipgloss.NewStyle().Foreground(lipgloss.Color(colour)).Render(ln))
+		}
+		out = append(out, "")
+	}
+	if found == 0 {
+		out = append(out, mutedStyle2().Render(" no review recorded for this task yet"))
+		out = append(out, "")
+		out = append(out, dimStyle().Render(" run one with:  bin/review "+m.task.Identifier+" --post"))
+	}
+	return out
+}
+
 func (m workModel) factsLines(width int) []string {
 	var out []string
 	out = append(out, "")