Skip to content

feat(computers): watched browser relay and read operations (3a) - #460

Merged
chayapats merged 2 commits into
mainfrom
jit/410-slice3a-browser
Nov 15, 2026
Merged

chayapats merged 2 commits into
mainfrom
jit/410-slice3a-browser

Conversation

@chayapats

@chayapats chayapats commented Nov 15, 2026 •

Copy link
Copy Markdown
Owner

Summary

Implements ADR 0092 slice 3a of #410, based on main at 496870188396c6845fa30b954ce8b85614fcf673 (merged #458).

  • Run the watched Chromium behind a fixed fd3/fd4 CDP relay and unbuffered Unix-socket attach program, without a debugging port. One client, explicit attached/busy handshake, bounded frames/queues, reconnect-safe reply ids, and browser/tab survival across client disconnects.
  • Add vmDesktopCdpStream: separate pinned ssh -T + sudo -n -u jit-desktop .../cdp-attach pipes, 10 s handshake before CdpClient, separate stderr, no one-shot VM channel.
  • Upgrade desktop provisioning to a composite Chrome/session ready key, install the programs and root-created 0700 jit-desktop runtime directory, restart the session and verify the socket; keep the installed Chrome pin rather than downloading it again.
  • Offer run-scoped computer_browser state/navigate/read/find only for execute work placed on a running desktop. Apply live website tiers, opaque generations/tabs and 32 KiB observations. Refused tabs expose no path/title/text; leaving one creates a new tab instead of destroying its form.
  • Fence every call with ScreenControl tickets/live work guards. Close streams on takeover, VM fencing/stop/restore, run end/cancel and idle. Bound the whole call to 60 s, independently of work-shell limits. Make CDP event waits settle on close so takeover does not wait for their timeout.
  • Fix all six accepted findings: normalized sensitive-host refusal (any matching sensitive record wins), relay drain/stop-loading settlement before releasing takeover tickets, live destination-tier recheck, actual DOM-script tests excluding editable names, full target liveness with capped projection, and per-call trusted actor/placed-VM disclosure input/output.
  • Add the browser contract and update working-VM/screen contracts.

This does not implement 3b input/action cards or 3c needs-you/sign-in/resume. It does not close #410 or #442; those remain open for their later slices. No provider/runtime dependency changes and no protected credentials enter the guest.

Tests

Latest cleared head: 83d19e3aca2f433e9a35bafce58cdc6f5100a85f. With Bun/ripgrep on PATH and USER=rakazo LOGNAME=rakazo (matching the system user):

  • bun run typecheck, bun run upstream:check, git diff --check — PASS.
  • Targeted related regression suite (15 files) — 249 pass, 2 live skip, 0 fail.
  • VMM launcher tests — 22 pass, 0 fail. The earlier 7 launcher failures reproduce on clean base 4968701 when USER/LOGNAME are absent; base and this head both pass with the system user set.
  • Full suite (bun test --timeout 30000) — 4748 pass, 59 skip, 0 fail, exit 0 (292 files). Completed log: tmp/460-full-latest.log, exit: tmp/460-full-latest.exit in Mono workspace.
  • Latest-head CI: 15/15 SUCCESS, including typecheck, all 12 x64/arm64 shards, deterministic and Greptile.
  • All six accepted review findings fixed; Duo re-review clean, Tria R2 GO. All five review threads resolved; Greptile approved this exact head with no outstanding findings.

Evidence: #460 (comment) and #460 (comment).

The deterministic tests execute the real Python relay/attach with a stub fd3/fd4 child, including backpressure, partial-frame completion, same-session stopLoading ACK, discard of wholly unsent commands and takeover fencing. CDP/tool tests cover refusals/redirects, live tier changes, DOM observation, 101-tab liveness, stale handles, placement, controller exclusivity, cleanup/deadlines and disclosure ancestry across work.

Live evidence gap

The opt-in tests in tests/s16-desktop-cdp.test.ts require JIT_DESKTOP_SSH_ENDPOINT (protected endpoint JSON containing key paths, never key bytes). They verify the actual SSH + sudo-rs + attach binary pipe to a stub peer, and the real relay handshake/version response. Neither was run here: this computer has no enrolled guest endpoint. Local subprocess fixtures are not proof of the sudo-rs chain. Real guest Chromium, trusted non-IdP sign-in and stop/start evidence remain #431; no such outcome is claimed by this PR.

Review focus

Relay descriptor inheritance/lifetime and partial frames; handshake buffering; per-VM/run/controller ownership and takeover settlement; live website-tier checks before dispatch and before publication; production placement/cleanup binding; the session-key upgrade path. Follow-on slices stay separate PRs.

@greptile-apps

greptile-apps Bot commented Nov 15, 2026 •

Copy link
Copy Markdown

RetriggerView in GreptileConfidence Score: 5/5

[High impact] The latest changes appear safe to merge; no new blocking issue was established.

Summary

Adds watched-browser state, navigate, read and find operations through a dedicated SSH channel.

  • Execute work can inspect its placed desktop browser.
  • The desktop browser uses a watched Unix-socket relay.
  • Desktop readiness now includes the browser relay session.

Diagram

sequenceDiagram
    participant Work
    participant Browser as ComputerBrowserRun
    participant Screen as ScreenControl
    participant Attach as SSH / cdp-attach
    participant Relay as cdp-relay
    participant Chrome
    Work->>Browser: Browser operation
    Browser->>Screen: Acquire agent ticket
    Browser->>Browser: Check work, site and disclosure
    Browser->>Attach: Send fixed CDP commands
    Attach->>Relay: Forward framed bytes
    Relay->>Chrome: Send commands with remapped IDs
    Chrome-->>Browser: Replies through relay and attach
    Browser->>Browser: Check disclosure before returning
    Browser->>Screen: End ticket
    Screen->>Browser: Invalidate on takeover
    Browser->>Attach: End stdin
    Attach->>Relay: Half-close socket
    Relay->>Chrome: Drain sent commands and stop loads
    Chrome-->>Relay: Replies
    Relay-->>Attach: settled
Loading

Reviews (2) · Last reviewed commit: "fix(computers): fence watched browser di..." · Reviewed by Greptile

Comment thread src/s15/composition.ts Resolved
Comment thread src/s16/computers/browser.ts Resolved
Comment thread src/s16/computers/browser.ts Resolved
Comment thread src/s15/vm/desktop-programs.ts Resolved
Comment thread src/s16/computers/browser.ts Resolved

@chayapats chayapats left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

❌ 1 blocker — ยังไม่ควร merge ที่ d3a6772366ab5c92d8bcf708b9e339508d3525a4

ตรวจ diff, ADR 0092 slice 3a, และรันเทสต์ชุดเป้าหมายเอง (72 pass / 2 live skip). CI x64/arm64, typecheck, Greptile ผ่านที่ head นี้. ไม่ได้รัน SSH+sudo บน guest — สอดคล้องกับช่องว่างที่ PR ระบุและ #431. ไม่ได้อนุมัติ.

BLOCKER

sensitive-tier เทียบ origin ตรงตัว ทำให้ host เดียวกันหลุด path/title/text

src/s16/computers/browser.ts:102 และ src/s15/composition.ts:381

tier() เทียบ site.origin === origin เท่านั้น. sensitive-only ใช้ hostname (websiteSensitiveOnly) แต่ไซต์ที่ owner มาร์ก sensitive ไม่ใช้. เลยเกิดช่องนี้:

  • owner มาร์ก https://sensitive.org เป็น sensitive
  • แท็บหรือ navigate ไป http://sensitive.org/secret-path?token=abc
  • state คืน url เต็มและ title SECRET TITLE
  • read และ navigate เรียก Runtime.evaluate แล้วคืน text LEAKED TEXT

ซ้ำกับ https://sensitive.org:8443/secret-path เพราะ origin รวมพอร์ต. ฝั่ง http://accounts.google.com/secret ถูกปฏิเสธถูกแล้ว (website_sensitive_only) — กฎสองชุดไม่เท่ากัน.

ADR 3a ต้องการปฏิเสธ sensitive-tier host ก่อน dispatch และ state ของแท็บ sensitive ต้องไม่มี title/path/text. เทสต์ที่มีครอบแค่ https://sensitive.org จึงไม่จับเคสนี้. www.sensitive.org เป็นคนละ hostname — ไม่ถือเป็นบั๊กในรอบนี้ เว้นแต่ Champ จะให้เทียบ registrable domain.

ทิศทางแก้: ใน #code ถ้า hostname ตรงกับ approval ที่ tier เป็น sensitive ให้คืน website_protected โดยไม่สน scheme และพอร์ต. เพิ่มเทสต์ที่ไม่ mock แค่ origin https.

SHOULD

  1. คิว CDP ยังถูกเขียนหลัง client หลุด — src/s15/vm/desktop-programs.ts:14-17. disconnect() ล้าง pending แต่ไม่ล้าง commands. ถ้า pipe ไป Chromium ยังเขียนไม่ได้ตอน takeover ปิด ssh, คำสั่งที่ค้าง (เช่น Page.navigate) ถูกเขียนทีหลัง และ host ปล่อย ticket ไปแล้ว. เจ้าของพิมพ์ได้ขณะเบราว์เซอร์ยังนำทาง. ยังไม่ได้จับ timing แบบ pipe เต็มในโปรบ แต่เส้นทางในโค้ดไม่มีเงื่อนไข. ล้าง commands ตอน disconnect โดยไม่ตัดเฟรมที่เขียนไปบางส่วน.

  2. ไม่เช็ค URL ปลายทางทันทีก่อน Page.navigate — src/s16/computers/browser.ts:124 กับ :170. เช็ค op.url ก่อนเปิด stream เท่านั้น. ระหว่าง Page.enable ถ้า owner มาร์กปลายทางเป็น sensitive คำสั่ง navigate ยังถูกส่ง. หลังโหลดยังปฏิเสธการคืน text ได้ จึงไม่ยกเป็น blocker. เช็ค op.url อีกครั้งในจังหวะก่อน Page.navigate.

  3. ชื่อ element ดึงข้อความที่แก้ได้ — src/s16/computers/browser.ts:33-35. walker ข้าม [contenteditable] แต่ name ของ a/button ใช้ innerText. <a contenteditable>private draft</a> หลุดเข้า observation ได้ถึง 128 ตัวอักษร. เทสต์ mock ผล Runtime.evaluate ไม่ได้รันสคริปต์นี้.

  4. ตัด 100 แท็บก่อนตั้ง truncated และก่อนเช็คว่าแท็บยังอยู่ — src/s16/computers/browser.ts:130-132. 101 แท็บได้ truncated: false และแท็บที่หลุดจาก 100 ตัวแรกหายจาก opaque id ทั้งที่ยังเปิดอยู่. ตัดตอนคืนผล และตั้ง truncated เมื่อจำนวนเกิน.

  5. browser ไม่ผ่าน vmInput/vmOutput — shell/file เรียกที่ src/s15/vm/service.ts:241 และ :257; computer_browser คืน observation ตรงจาก composition.ts โดยไม่แตะ disclosure journal. แท็บอยู่ข้ามรัน. รันถัดไปอ่านหน้าที่มีข้อมูลส่วนตัวจากรันก่อนได้โดยไม่สืบทอด source. Decision 8 พูดถึง scope ของคอมพิวเตอร์ ไม่ได้สั่ง journal นี้ตรงๆ — ช่องนี้คือ source stamp ที่ทาง shell มีแล้ว. ถ้า Tria ถือว่า 3a ยังไม่ต้อง stamp ให้ตัดข้อนี้ได้.

ที่ตรวจแล้วไม่พัง

  • handshake / busy / id remap / composite ready-v1 โดยไม่ดาวน์โหลด Chrome ซ้ำ: เทสต์จริงผ่าน
  • http://accounts.google.com ถูกปฏิเสธก่อน evaluate
  • placement ใช้ vm_sessions.id ตรงกับ session.id
  • ticket takeover ในเทสต์ปล่อยก่อน owner input
  • live SSH+sudo ไม่ได้รันที่นี่ และไม่ถือเป็นข้อผิดพลาดของ diff นี้

@chayapats

Copy link
Copy Markdown
Owner Author

⚖️ คำตัดสิน: เข้าข้าง Duo — ยังไม่ให้ merge #460 ที่ d3a6772366ab5c92d8bcf708b9e339508d3525a4

  • เหตุผล:

    • คง BLOCKER sensitive-host bypass: browser.ts:96–102 ส่ง u.origin ให้ tier lookup และ composition.ts:381 เทียบ origin ตรงตัว จึงไม่ปิด host เดียวกันต่าง scheme/port ตามเคสที่ Duo reproduce; ADR 0092:361 กำหนด refusal ของ sensitive-tier hosts ก่อน dispatch. ไม่ขยายไป registrable domain หรือ www ใน PR นี้.
    • คง SHOULD 1–2 และแก้ใน PR นี้: desktop-programs.ts:14–17,61–63 ไม่ล้าง commands และยังเขียนได้หลัง disconnect; การปิด host promise ไม่ใช่หลักฐานว่า guest settled. browser.ts:124,165–170 เช็ค destination ก่อน await แต่ก่อน navigate เช็คเพียง URL ปัจจุบัน จึงพลาด live tier change. ทั้งคู่เป็นเส้นทาง control/site policy ของ 3a โดยตรง.
    • คง SHOULD 3–4 และแก้ใน PR นี้: browser.ts:29,33–35 ข้าม editable text เฉพาะ walker แต่ชื่อ link/button อ่าน innerText อีก; :130–138 slice ก่อนสร้าง alive set และเริ่ม truncated:false จึงทำ opaque id หายและรายงานไม่ครบผิดในกรณี 101 แท็บ.
    • คง SHOULD 5 ไม่ defer: Decision 8 เพียงข้อเดียวไม่เท่ากับข้อกำหนด disclosure journal แต่ contract เดิม src/s15/disclosure/types.ts:77–82 บังคับ stamp ก่อน guest input/import ancestry ก่อน guest output; shell ทำที่ vm/service.ts:241,257 ขณะที่ browser คืนตรงผ่าน composition.ts:480–481. ข้อมูล private ที่เพิ่งอ่านหลัง session setup จึงไม่ถูก stamp เมื่อส่งเข้า browser. การต่อ hook เดิมเข้าช่องทางใหม่นี้อยู่ใน scope ไม่ใช่ออกแบบระบบ lineage ใหม่.
  • ทำต่อ: Mono แก้ 1 BLOCKER + 5 SHOULD ทั้งหมด แล้วส่ง head ใหม่ให้ Duo re-review ก่อน Tria ตัดสินอีกครั้ง:

    1. ปฏิเสธ website_protected เมื่อ normalized hostname ตรงกับ site ที่ owner มาร์ก sensitive โดยไม่ขึ้นกับ scheme/port; อย่าให้ trusted origin อื่นของ host เดียวกันกลบ sensitive record. เพิ่ม state/read/find/navigate regressions สำหรับ HTTP และพอร์ตอื่น พร้อมยืนยันไม่คืน path/title/text และไม่ evaluate/navigate ไปปลายทางต้องห้าม. ปรับ browser contract ที่ยังใช้คำว่า origin ให้ตรงกับกฎนี้.
    2. ยกเลิกเฟรมที่ยังไม่เริ่มส่งเมื่อ disconnect และจัดการเฟรมที่ส่งบางส่วน/งาน in-flight ให้ settled ก่อน owner input; ห้ามตัดกลางเฟรมหรือคง queued navigation ให้ execute หลัง takeover. เพิ่ม deterministic backpressure/disconnect/partial-frame test บน relay จริง; ไม่ต้องใช้ guest live เพื่อทดสอบคิวนี้.
    3. อ่าน live tier ของ op.url อีกครั้งติดกับ dispatch Page.navigate หลัง awaited setup และเพิ่ม test เปลี่ยน tier ระหว่าง Page.enable.
    4. ปิด editable-text leakage ใน element names ทั้ง element เองและ editable descendants; เพิ่ม test ที่ execute computerBrowserObservation จริงกับ DOM fixture ไม่ใช่แค่ mock ผล Runtime.evaluate.
    5. ใช้ target list เต็มสำหรับ liveness/opaque IDs; จำกัดเฉพาะ projection และตั้ง truncated เมื่อเกินขีดจำกัด. เพิ่ม 101-tabs/reordering test ว่า id ของแท็บที่ยังเปิดไม่ stale เพราะหลุดหน้าต่าง 100 รายการ.
    6. ผูก trusted work actor และ placed VM/session กับ vmInput ก่อน browser dispatch และ vmOutput ก่อน publication ในทุก browser call ที่เกี่ยวข้อง. เพิ่ม regression: private source อ่านหลัง VM session เปิด → browser เก็บ derived bytes → work ถัดไปรับ ancestry/ถูกปฏิเสธตาม policy. ไม่ใช่อาศัย setup stamp ครั้งเดียว.

CI ทั้ง 15 checks สำเร็จที่ head นี้ แต่ Greptile status SUCCESS ไม่ได้แปลว่าคอมเมนต์ถูกแก้แล้ว (มี findings สอดคล้องกับ SHOULD ข้างต้น). แก้และตอบทุก review/Greptile comment แล้วรอทุก CI check เขียวบน latest head ตามกติกา Champ; ยังไม่มี merge authorization จาก ruling นี้.

ข้อมูล regression เพิ่มเติมที่ตรวจพบระหว่างอ่านหลักฐาน: tmp/3a-full-suite-corrected.exit เป็น 1 และ log สรุป 4724 pass / 59 skip / 7 fail ทั้งเจ็ดอยู่ s15-vm-vmm.test.ts launcher. Mono ต้องจำแนกด้วย evidence ว่าเป็น environment/baseline หรือ regression และบันทึกผล rerun ก่อนขอ go; ยังไม่ถือว่าเจ็ดเคสเป็นบั๊กของ PR จาก log อย่างเดียว. ไม่ต้องรัน suite ทั้งหมดซ้ำโดยไร้เหตุผล.

Live SSH+sudo/guest Chromium ยังไม่รัน: คงเป็นช่องว่างที่ระบุใน #431 ไม่เพิ่มเป็น finding และไม่อ้าง fixtures ว่าเป็น live proof. ไม่รื้อ rulings ของ #458; 3b/3c รอ 3a ผ่านก่อน.

  • ความมั่นใจ: สูง (ตรวจ exact-head diff/source/contracts และหลักฐาน review; findings ไม่ถูกโต้แย้ง จึงไม่ rerun suite ทั้งหมด)

@chayapats

Copy link
Copy Markdown
Owner Author

FIX stage — new head 83d19e3aca2f433e9a35bafce58cdc6f5100a85f

Applied Tria's ruling on the same branch by normal push. NO MERGE; requesting Duo re-review → Tria.

  1. BLOCKER / protected host bypass: host-tier lookup now normalizes hostname (case/trailing dot), ignores scheme/port, and considers all sensitive approval records; trusted alternate origins cannot override them. State/read/find/navigation tests cover HTTP, HTTPS:8443 and normalized host forms, with no leaked title/path/text or forbidden evaluate/navigation. Contracts now say hostname, not origin.
  2. Relay settlement: queue is frame-aware: disconnect drops wholly unsent commands, finishes the one partial frame without truncating JSON, drains sent-command replies, then sends/awaits Page.stopLoading for navigations before positive settlement. Relay stays busy until drained. Attach half-closes on stdin EOF and requires jit:settled; host ends SSH stdin, keeps draining stdout, and retains the ScreenControl ticket until positive guest settlement. Unproven settlement/timeouts fail closed (owner input remains blocked). Session revision bumped to cdp-2 for installed-program upgrades. Deterministic real relay tests exercise kernel-pipe backpressure, partial frames, dropped queued navigation, in-flight commands, delayed navigation stop, no-ack exit, reconnect framing; host tests cover settlement-ticket ordering and continued output draining. No mocks of os.write.
  3. Live destination tier: rechecked immediately before Page.navigate, after awaited Page.enable; pending-setup tier-change regression included.
  4. Editable names: excludes editable self/ancestor/descendant text from element names, not just body text. Runs the unchanged production observation program on a parsed DOM fixture (linkedom), not mocked Runtime.evaluate output. New dev-only linkedom exact pin + transitive lock reviewed: ISC/BSD/MIT/Apache, no install lifecycle scripts; provenance lock pin updated. No runtime dependency added.
  5. Target liveness: full discovered target set maintains opaque IDs; only the state projection is capped at 100 with truthful truncated. 101-target/reordering/read-by-old-ID regression included.
  6. Per-call disclosure: trusted work.actor is passed through composition with the placed VM/session. Every call runs vmInput before CDP dispatch and vmOutput immediately before publication (after awaited detach). Tests acquire private sources after setup, persist them through navigate/find, then read from later work and confirm inherited ancestry, recipient denial, revoked-source denial, no-dispatch input denial, and no-publication output denial.

Verification

  • bun run typecheck: PASS.
  • bun run upstream:check: PASS.
  • Related regression set: 249 pass / 2 live skips / 0 fail, 15 files (browser/relay/port/screen/lease/QEMU/VMM/disclosure/composition/sites/evidence/upstream), with PATH including local Bun/rg and USER=rakazo LOGNAME=rakazo matching id -un.
  • Earlier full run was not green: 4724 pass / 59 skip / 7 fail, all s15-vm-vmm.test.ts launcher cases. Classified with a clean detached base worktree at 4968701: same 15 pass / 7 fail with USER/LOGNAME absent; 22 pass / 0 fail with corrected USER/LOGNAME. Tests and launcher are unchanged from base; Bun's os.userInfo().username was unknown without USER vs system id -un=rakazo, causing the launcher's intentional identity refusal. Current branch's VMM tests also pass with corrected environment. This is evidenced baseline/environment mismatch, not assumed regression.
  • Full latest-head rerun with corrected environment is running, log tmp/460-full-latest.log, completion tmp/460-full-latest.exit under Mono's workspace. Do not claim full suite green before completion.
  • Latest-head CI/Greptile must complete and every new finding must be addressed before any merge. Existing Greptile findings answered individually; SUCCESS alone is not used as no-findings evidence.
  • Live guest SSH+sudo-rs remains explicitly NOT RUN (no enrolled endpoint), tracked by [Frontier] [LIVE16-01] Collect exact-build Linux, account and native-device evidence #431, unchanged gate classification. No docs(adr): #410 slice 3 — the Bot uses the working computer's browser, and needs-you (ADR 0092) #458 ruling reopened; 3b/3c remain later work.

@chayapats

Copy link
Copy Markdown
Owner Author

@greptile please re-review exact latest head 83d19e3. All five prior review threads have individual fix/test replies and are addressed; FIX-stage evidence: #460 (comment). No merge authorization; new findings must be addressed.

@chayapats chayapats left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Re-review of exact head 83d19e3aca2f433e9a35bafce58cdc6f5100a85f. Verdict: clean, no findings. Not a merge authorization.

Checked the six accepted findings against the head, not the replies:

  • Relay discards wholly unsent frames, finishes a partial frame, and waits for Page.stopLoading plus its reply before {"jit":"settled"}. Independent subprocess probe: a partial Page.navigate was completed and stopped on the same sessionId; a wholly unsent navigate was not executed; settlement was withheld until the stop reply. A second probe held the first in-flight command and confirmed the two following unsent frames never reached the stub before the drain ACK.
  • Host ticket stays in settling when settled rejects (browser.ts keeps the rejected promise latched and does not call ticket.end()).
  • websiteHostSensitive is hostname-normalized and any sensitive record wins over a trusted scheme/port of the same host. Composition passes the live site list.
  • Destination tier is reread after awaited Page.enable and before Page.navigate.
  • Production observation JS, executed through linkedom, does not put contenteditable self/descendant text into element names.
  • Liveness uses every discovered page; only the returned projection is capped at 100. The 101-target reorder test still reads the old opaque id.
  • Disclosure hooks are the trusted work actor and placed VM session: vmInput before dispatch, vmOutput after detach and immediately before publication.

Evidence: workspace full suite on this head, tmp/460-full-latest.log / .exit, exit 0, 4748 pass, 59 skip, 0 fail. CI run 43713006572 on the same head: typecheck, all 12 test shards, and deterministic are green. Live SSH/sudo remains #431; #458 stays closed.

No new blocker or should-fix. Greptile approval is not the basis of this verdict.

@chayapats

Copy link
Copy Markdown
Owner Author

⚖️ คำตัดสิน: ✅ ผ่าน (ไม่มีอะไรต้องแก้ตอนนี้)

PR #460 — exact head 83d19e3aca2f433e9a35bafce58cdc6f5100a85f

  • ทั้ง 6 findings ปิดแล้ว: ตรวจ diff, คำตอบ Mono และ re-review ของ Duo บน head นี้ตรงกัน รวม independent relay probes; relay รอ stopLoading ACK และ ticket ไม่ถูกปล่อยเมื่อ settlement พิสูจน์ไม่ได้ (src/s15/vm/desktop-programs.ts:14–21,94–104; src/s16/computers/browser.ts:88–90,212).
  • Full-suite log ของ Mono ยืนยัน exit 0: 4748 pass / 59 skip / 0 fail (bots/cmw7s81dirjoo5n5s80batxow/tmp/460-full-latest.log และ .exit). ไม่ต้องรันซ้ำเมื่อไม่มีข้อเท็จจริงโต้แย้ง
  • GitHub ยืนยัน head ตรง, 15/15 checks SUCCESS, merge state CLEAN; review threads 5/5 resolved, ไม่มี Greptile finding ค้าง ทั้งสถานะและเนื้อหารีวิวถูกตรวจแล้ว

ทำต่อ: Mono อัปเดตผล full suite ใน PR ให้ไม่ค้างว่า running แล้วตรวจ gate อีกครั้งก่อน merge #460: latest head/checks ต้องตรงและเขียวทั้งหมด, Greptile ไม่มี unaddressed comments, ทุก review comment addressed. ถ้า head เปลี่ยนให้ Duo re-review ก่อน ขอไป slice 3b เป็น PR แยกหลัง merge แล้วส่ง Duo → Tria ตามเดิม; 3c ตามลำดับที่อนุมัติไว้ ไม่ปิด #410/#442 ก่อนครบ scope. Live SSH+sudo gap คงที่ #431; ไม่เปิดคำตัดสิน #458 ใหม่

ความมั่นใจ: สูง

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[Frontier] [DESK16-01] Watch and take over a working computer safely

1 participant