feat: pick a turn a usage limit stopped back up once the limit resets

CHANGELOG.md

@@ -45,6 +45,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   BMP, ICO and PDF up to 20 MB per side; the Pulls screen still says "Binary
   file".
 
+- **A turn your plan's usage limit stopped picks up again once the limit
+  resets.** On Claude Code and Codex, the card says which limit stopped the
+  session and when it lifts, and lich types a prompt asking the agent to
+  continue two minutes after the reset. Any turn that starts before then, yours
+  or the provider's own, cancels it, and so does **Scheduled…** › **Cancel it**
+  in the card's menu. Nothing is parked for a limit that resets more than a day
+  out. Claude Code needs the lich-plugin release that reports `StopFailure`.
+
 ### Changed
 
 - **The window is laid out as islands.** The session sidebar, the terminals and
@@ -62,6 +70,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   worktree's long path no longer stretches the cards wider than the sidebar and
   pushes the rest of the list sideways while the drag is in flight.
 
+- **A Codex card no longer spins forever after a usage limit ends its turn.**
+  Codex reports no end for that turn, so lich now reads it off the session's own
+  record.
+
 ## [0.61.0] - 2026-10-07
 
 ### Added

README.md

@@ -100,6 +100,7 @@ reads what each CLI writes down and no two of them write down the same things.
 | Context window in the footer | yes | yes | no | no | no | no | no | yes |
 | Cost in the footer | yes | yes | no | yes | yes | yes | no | credits |
 | How much of your plan is left | yes | yes | no | no | no | no | no | no |
+| A turn your plan's usage limit stopped picks up again once the limit resets | yes | yes | no | its own retries | its own retries | no | no | no |
 | Spinner while a turn runs, ring when it ends | yes | yes | yes | yes | yes | no | no | yes |
 | Bell when the agent is blocked on you | yes | yes | no | yes | no | no | no | no |
 | Sidebar filter by state: Waiting, Running, Unread | yes | yes | no Waiting | yes | no Waiting | no, always Idle | no, always Idle | no Waiting |

README.zh-CN.md

@@ -86,6 +86,7 @@
 | 底栏里的上下文窗口 | 有 | 有 | 无 | 无 | 无 | 无 | 无 | 有 |
 | 底栏里的花费 | 有 | 有 | 无 | 有 | 有 | 有 | 无 | 按点数计 |
 | 套餐额度还剩多少 | 有 | 有 | 无 | 无 | 无 | 无 | 无 | 无 |
+| 被套餐用量上限打断的一轮，在上限重置后自动接着做 | 有 | 有 | 无 | 由其自身重试 | 由其自身重试 | 无 | 无 | 无 |
 | 一轮进行中转圈，结束后成环 | 有 | 有 | 有 | 有 | 有 | 无 | 无 | 有 |
 | 智能体卡在等你时响铃 | 有 | 有 | 无 | 有 | 无 | 无 | 无 | 无 |
 | 侧栏按状态筛选：Waiting、Running、Unread | 有 | 有 | 无 Waiting | 有 | 无 Waiting | 无，总在 Idle | 无，总在 Idle | 无 Waiting |

docs/ceilings.md

@@ -580,6 +580,21 @@ work when nobody knows it and that the call site never shows. The mechanism and
   those logins and never writes them: it does not refresh the token, so an expired one reads as signed out until
   the provider's own CLI rotates it. A reading is cached for five minutes because both endpoints rate-limit hard —
   the number on screen is up to that old, and nothing on it says so.
+- **A turn a usage limit stopped is picked back up on two providers only** (`internal/terminal/limit.go`,
+  `internal/relay/resume.go`): Claude Code and Codex are the two that write the limit down with its reset. On
+  Antigravity, Crush, Cursor CLI and Kiro CLI nothing marks a limit at all, so the card keeps spinning until the
+  next prompt; opencode and oh-my-pi retry by themselves and lich adds nothing on top. Claude Code's end arrives
+  only as `StopFailure`, which an older lich-plugin does not report. Codex raises no hook for it, so an open
+  Codex turn is read off its rollout every 30 seconds, and the limit is *inferred*: a turn completing with
+  nothing said right after a window shows 100% spent (measured on `codex exec` 0.144.5, never on the TUI); a
+  change to that shape stops the read without a word. The continuation is a scheduled prompt told apart from a
+  person's by its exact text, so someone parking those same words has them dropped on the session's next turn.
+  Nothing is parked for a reset already past or more than 24 hours out (Claude Code's own auto-continue gives up
+  at the same horizon), nor over a prompt already parked there. Any turn starting before it is due drops it,
+  which is also what keeps it from landing behind Claude Code's own `autoContinueAtUsageLimit`, a setting the
+  server enables per account, which is why the prompt is due two minutes after the reset rather than at it. The
+  card's limit rung lives in the window alone: a reload drops it, while the parked prompt stays on the card as a
+  plain scheduled one.
 - **The pace marker is borrowed calibration, and it is silent far more often than it is wrong**
   (`internal/quota/pace.go`): the two numbers that decide when a weekly window is marked as spending ahead —
   fifteen percentage points past the elapsed share, and no marking at all in the first twenty-four hours after

docs/hooks/session-state.md

@@ -48,7 +48,7 @@ Both sides test against the payloads in
 | `PreToolUse`       | `PreToolUse`        | `PreToolUse`     | `tool.execute.before`    | `tool_call`    | —          | dropped            | `preToolUse`       | `busy` + `tool` |
 | `PostToolUse`      | `PostToolUse`       | —                | `tool.execute.after`     | `turn_start`   | —          | dropped            | `postToolUse`      | `busy`    |
 | `Notification`     | `PermissionRequest` | —                | any `*.asked`            | —              | —          | —                  | —                  | `waiting` + `reason` |
-| `Stop`             | `Stop`              | `Stop`           | `session.status` (`idle`) | `session_stop` | —          | —                  | `stop`             | `done`    |
+| `Stop`, `StopFailure` | `Stop`              | `Stop`           | `session.status` (`idle`) | `session_stop` | —          | —                  | `stop`             | `done`    |
 | `SessionEnd`       | —                   | —                | —                        | —              | —          | `SessionEnd`       | —                  | `idle`    |
 
 **Kiro closes four of the six rows and neither of the other two.** It has no
@@ -69,6 +69,20 @@ same payload. Its `idle` means the turn ended, which is lich's `done` — not
 lich's `idle`, which says the CLI itself has left. Nothing in opencode's event
 list says that, so like Codex it never reports `idle`.
 
+**Claude Code ends a turn an API error stopped with `StopFailure`, not `Stop`.**
+It fires instead of `Stop` when a usage limit, an overload or a failed login ends
+the turn (2.1.293), and a client that registers only `Stop` leaves the card
+spinning until the next prompt. It reports `done` like `Stop`: the turn is over
+either way. Which error it was is not in the report: lich reads it off the
+transcript on every `done`, and a usage limit it finds there parks the turn's
+continuation for when the limit resets (docs/ceilings.md).
+
+**Codex reports no end at all for a turn a usage limit stopped**: measured on
+0.144.5, `UserPromptSubmit` fires and `Stop` never does. lich ends that turn
+itself, reading the rollout of every open Codex turn, and publishes it as
+`interrupted`, the state it already raises for a turn that ended without the
+provider saying so.
+
 **oh-my-pi carries `busy` on `turn_start` rather than after a tool.** Its
 non-interactive runs never emit `input` at all, so the turn boundary is the one
 event every turn passes through — the same place the title is re-read, since omp

frontend/src/components/render-budget.test.tsx

@@ -238,10 +238,11 @@ async function mountSidebar() {
   return budget
 }
 
-// What one card drags along when it repaints: its own chrome, and the base-ui
-// context-menu and tooltip parts wrapped around it. The unnamed pair are
-// base-ui's own anonymous render functions.
+// What one card drags along when it repaints: its own chrome, its status line,
+// and the base-ui context-menu and tooltip parts wrapped around it. The unnamed
+// pair are base-ui's own anonymous render functions.
 const CARD_CHROME = {
+  SessionStatusRung: 1,
   ContextMenu: 1,
   ContextMenuRoot: 1,
   ContextMenuTrigger: 2,
@@ -285,10 +286,12 @@ test("a finished turn repaints that session's card and no other block", async ()
   await budget.unmount()
 })
 
-test("a todo event repaints that session's card and nothing else", async () => {
+// The task list is read by the status line alone, so its count moving repaints
+// that line and leaves the card around it, and its menus and tooltip, alone.
+test("a todo event repaints that session's status line and nothing else", async () => {
   const budget = await mountSidebar()
   await budget.act(() => bus.emit(TODO_EVENT, { id: "s2", done: 3, total: 7 }))
-  expect(budget.take()).toEqual({ "SessionCard#s2": 1, ...CARD_CHROME })
+  expect(budget.take()).toEqual({ SessionStatusRung: 1 })
   await budget.unmount()
 })