Fix A run's agent session that

- signoff (visit 1): pass, then rebase: approved by rule autonomous
- rebase (visit 1): pass, then wait-rebased: exit 0
- wait-rebased (visit 1): pass, then rebased-green: held after 40 checks
- rebased-green (visit 1): pass, then land: exit 0
- land (visit 1): pass, then wait-landed: exit 0
- wait-landed (visit 1): pass, then landed-green: held after 54 checks
- landed-green (visit 1): fail, then end: exit 1
    PR #300 at 03a7618: failed: check; failing tests: TestScripts/describe-golden
- landed-green (visit 2, retried): fail, then end: exit 1
    PR #300 at 0296e5a: not settled: 0 of 1 checks finished
- landed-green (visit 3, retried): pass, then rebase-merge: exit 0
- rebase-merge (visit 1): pass, then wait-rebased-merge: exit 0
- wait-rebased-merge (visit 1): pass, then rebased-merge-green: held after 25 checks
- rebased-merge-green (visit 1): pass, then merge: exit 0
- merge (visit 1): pass, then wait-main: exit 0
- wait-main (visit 1): pass, then main-green: held after 1 checks
- main-green (visit 1): pass, then close-out: exit 0
- close-out (visit 1): pass, then ship: exit 0
- ship (visit 1): pass, then end: exit 0

## Leaves

- run-session-idle-stall (step build/build): passed: attempt 2: verifier passed
  - attempt 1 builder, session 97f9ddb2-d81e-4471-97f1-01a1aae0f209: pass (said done)
  - attempt 1 verifier, session e63658e6-8638-4d2e-b29b-01a1aae93459: fail (verifier found 1 problem(s))
  - attempt 2 builder, session 90748e6f-5f2b-44a9-b065-01a1aaf12a89: pass (said done)
  - attempt 2 verifier, session c99dfaa9-a4b6-4e02-9261-01a1aaf43b9a: pass (Evidence written by the verifier)
  - attempt 1 finding f1: Decision idle-stall-autonomous-only, which answers question review-01m6n9ad-r3-f1 that blocks this leaf, rules that the idle-without-done stall applies only to sessions an autonomous run launched (session.Record.Autonomous true): an idle session of an owner-attended run is never stalled, and the acceptance gains an idle record without done and with Autonomous false that is not stalled at T+21m. Commit 8a2422c4 ignores it: internal/session/status.go Stalled never reads r.Autonomous, so an attended idle session (e.g. a plan run's interview waiting on the owner) counts as stalled after 20 minutes and the owner's n would send it carryOn. TestStalled in internal/session/status_test.go has no Autonomous case (its idle-at-21m case passes with Autonomous unset, i.e. attended), and inbox-stalled.txtar has no attended-idle case. The leaf's literal acceptance passes (go test -trimpath ./internal/session -run TestStalled, ./internal/api -run TestRunStalledNow, ./cmd/rhiza -run TestScripts/inbox-stalled all PASS), but the leaf's statement/acceptance were never amended as the Decision requires; the fix is to stall idle only when r.Autonomous, with tests for an attended idle record (not stalled at 21m) and an autonomous one (stalled), and to amend the leaf's acceptance.
- daemon-nudges-stalled (step build/build): passed: attempt 2: verifier passed
  - attempt 1 builder, session 324cafb8-ef54-449a-9f5e-01a1aaf65da6: pass (said done)
  - attempt 1 verifier, session 039b8091-fd53-46cf-bdbf-01a1aaf95574: fail (verifier found 1 problem(s))
  - attempt 2 builder, session 7100d12a-33dc-4644-9d2f-01a1aafa432e: pass (said done)
  - attempt 2 verifier, session ab6612fb-e5e3-4058-8064-01a1aafb0695: pass (Evidence written by the verifier)
  - attempt 1 finding f1: Acceptance requires the Go test to show the stalled autonomous session 'gets one message naming background shells and `rhiza session done`', but TestDaemonNudgeStalled (internal/api/daemon_nudge_test.go) never sees or asserts the message text: daemonNudgeStalled's send callback takes only the session ID, and the text (daemonNudgeText) is bound in daemonNudge, which no test covers. The other points pass (ran `go test -trimpath ./internal/api -run 'TestDaemonNudge|Stall' -count=1 -v`: PASS; one nudge per stall keyed on report time, no resend at the same stall, second nudge after a new report + 21 min, owner-attended run never nudged, inbox still lists both stalled sessions). Fix: have the fake sender receive the text (e.g. send(sid, text)) and assert it mentions a background shell and `rhiza session done`.
- daemon-nudges-stalled (step review/review): open
  - attempt 2 fixer in cycle 2, session b3855a86-caf8-4951-8db8-01a1ab009400: pass (said done)
  - attempt 2 verifier in cycle 2, session 4fbdf9ea-e661-450d-b29a-01a1ab012cf9: pass (Evidence written by the verifier)

## Checks

- compliance (step review/review): passed after 2 cycles
  - cycle 1: failed, check session 06f88b19-2d21-4df0-aed6-01a1aafc582c pass (said done), critic session 97d4d83e-6a0d-4d74-b41b-01a1aaffef3d pass (said done)
  - cycle 2: passed, check session 9fd71d47-30ac-42dd-8de4-01a1ab066fb7 pass (said done)
  - round 1 (cycle 1): 2 findings, 1 confirmed
    - f1 major at internal/api/daemon_nudge_test.go:22 on daemon-nudges-stalled: The acceptance requires 'a session of the autonomous run idle without done for 21 minutes gets one message'; the test writes every session with StateWorking (report() sets session.StateWorking for both 'auto' and 'owner'), so the idle-at-prompt case the change set exists for is never driven through the nudge. The leaf's own evidence concedes this. Make the autonomous session's hook report idle (its record is already Autonomous: true, so session.Stalled stalls it) and keep the asserted single message, no resend, second stall and inbox listing. — confirmed: The acceptance says the autonomous run's session is 'idle without done for 21 minutes'; TestDaemonNudgeStalled (internal/api/daemon_nudge_test.go:22) writes StateWorking for both records, so the idle stall this change set exists for (session.Stalled stalling idle only when Record.Autonomous) never reaches daemonNudgeStalled in any Go test. The acceptance is achievable as written: the 'auto' record can report idle (it is Autonomous) while the 'owner' record stays working, keeping both in the inbox and the attended one un-nudged, as the QA evidence did by hand. Not met as written; the fix is a one-line test change.
    - f2 minor at internal/api/daemon_nudge_test.go:47 on daemon-nudges-stalled: The acceptance names 'a Go test of the daemon poll'; the test calls daemonNudgeStalled directly with hand-built run.State values, so daemonNudge's run selection (live, not ended, journal Autonomous) and its call from runDaemon are only covered by reading the diff.
  - round 2 (cycle 2): no findings
- quality (step review/review): passed after 1 cycle
  - cycle 1: passed, check session 2e388159-1693-4ba3-87bb-01a1aafc7a36 pass (said done)
  - round 1 (cycle 1): 2 findings, 0 confirmed
    - f1 minor at internal/api/daemon_nudge.go:14 on daemon-nudges-stalled: daemonNudgeText opens 'You have been idle for over 20 minutes', but runStalledAll also returns sessions stalled while working (silent mid-turn); those get the same text, queued as a pending pointer, which misstates their state. Wording such as 'silent for over 20 minutes' covers both.
    - f2 nit at internal/api/run_test.go:842 on run-session-idle-stall: The 'idle' record gets a Done mark with the comment 'An idle session stalls too, unless it has said it is done', but the record is not Autonomous, so it would not stall without the mark either; the Done mark is redundant and the comment misdescribes why the session is skipped (an attended idle session never stalls).
- qa (step review/review): passed after 1 cycle
  - cycle 1: passed, check session 07b49e62-ff98-404d-af19-01a1aafcae45 pass (said done)
  - round 1 (cycle 1): no findings
- scope (step review/review): passed after 1 cycle
  - cycle 1: passed, check session e1c5ec31-eeaa-492a-8ee1-01a1aafcd502 pass (said done)
  - round 1 (cycle 1): no findings
- model (step review/review): passed after 1 cycle
  - cycle 1: passed, check session b2f53687-620c-49dd-b93b-01a1aafd0bd0 pass (said done)
  - round 1 (cycle 1): 4 findings, 0 confirmed
    - f1 minor at internal/api/daemon_nudge_test.go:21 on daemon-nudges-stalled: The leaf's acceptance says the autonomous run's session is 'idle without done for 21 minutes', but TestDaemonNudgeStalled writes StateWorking reports for both sessions; the evidence admits this. Either the acceptance should say 'stalled (working, or idle in an autonomous run)' or the test should use an idle autonomous record, so model and code say the same thing.
    - f2 minor at internal/api/daemon_nudge.go:14 on daemon-nudges-stalled: The capability says the nudge tells a stalled session to check background shells or say done; daemonNudgeText opens 'You have been idle for over 20 minutes', yet runStalledAll also returns working-stalled sessions (any run), which the daemon nudges too. The text claims idle where the model's stall covers working; word it as 'silent'.
    - f3 nit at .rhiza (decision idle-stall-autonomous-only): idle-stall-autonomous-only narrows run-session-idle-stall, run-status-stalled-session, inbox-stalled-items and idle-run-session-stalls, but has only an answers link to the review question; it has no decides link to the capabilities it shapes, so slicing those leaves does not surface it directly (it is only named in idle-run-session-stalls' text).
    - f4 nit at .rhiza (capability daemon): The daemon capability's statement (starts, resumes, reports) does not mention that it now nudges stalled sessions of autonomous runs; daemon-nudges-stalled sits under it, so this is only a summary gap.

## Code checks

- step review/code: pass

## Sign-offs

- signoff (visit 1): approved by rule autonomous

## Retries

- 2026-11-05T10:34:20Z retried at landed-green, after: step "landed-green" ended with outcome "fail"
- 2026-11-05T10:45:38Z retried at landed-green, after: step "landed-green" ended with outcome "fail"

## Sessions

- build/build, session 97f9ddb2-d81e-4471-97f1-01a1aae0f209: tokens input 30, output 5213, cache creation 36227, cache read 835081; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session e63658e6-8638-4d2e-b29b-01a1aae93459: tokens input 20, output 3287, cache creation 34907, cache read 542515; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session 90748e6f-5f2b-44a9-b065-01a1aaf12a89: tokens input 30, output 7375, cache creation 45874, cache read 938145; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session c99dfaa9-a4b6-4e02-9261-01a1aaf43b9a: tokens input 16, output 2893, cache creation 32683, cache read 426026; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session 324cafb8-ef54-449a-9f5e-01a1aaf65da6: tokens input 30, output 6900, cache creation 53739, cache read 998322; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session 039b8091-fd53-46cf-bdbf-01a1aaf95574: tokens input 16, output 2763, cache creation 29171, cache read 404391; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session 7100d12a-33dc-4644-9d2f-01a1aafa432e: tokens input 12, output 1782, cache creation 28405, cache read 293655; helper tokens input 0, output 0, cache creation 0, cache read 0
- build/build, session ab6612fb-e5e3-4058-8064-01a1aafb0695: tokens input 18, output 3581, cache creation 31025, cache read 468498; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 06f88b19-2d21-4df0-aed6-01a1aafc582c: tokens input 12, output 2816, cache creation 36319, cache read 319325; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 2e388159-1693-4ba3-87bb-01a1aafc7a36: tokens input 20, output 3626, cache creation 45000, cache read 617176; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 07b49e62-ff98-404d-af19-01a1aafcae45: tokens input 42, output 11849, cache creation 50555, cache read 1398213; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session e1c5ec31-eeaa-492a-8ee1-01a1aafcd502: tokens input 8, output 1350, cache creation 34156, cache read 190930; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session b2f53687-620c-49dd-b93b-01a1aafd0bd0: tokens input 18, output 4244, cache creation 49678, cache read 563413; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 97d4d83e-6a0d-4d74-b41b-01a1aaffef3d: tokens input 12, output 2061, cache creation 28312, cache read 296088; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session b3855a86-caf8-4951-8db8-01a1ab009400: tokens input 12, output 1639, cache creation 23313, cache read 281451; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 4fbdf9ea-e661-450d-b29a-01a1ab012cf9: tokens input 34, output 3336, cache creation 34255, cache read 812209; helper tokens input 0, output 0, cache creation 0, cache read 0
- review/review, session 9fd71d47-30ac-42dd-8de4-01a1ab066fb7: tokens input 10, output 1767, cache creation 37197, cache read 259805; helper tokens input 0, output 0, cache creation 0, cache read 0

Totals: tokens input 340, output 66482, cache creation 630816, cache read 9645243; helper tokens input 0, output 0, cache creation 0, cache read 0