You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Buildkite: the BUILDKITE_READ_TOKEN authenticates, but organizations/hunter-bown returns 404 No organization found and GraphQL viewer.organizations is empty. The token appears to have lost org access, possibly with the move to codewhale-hq. No PR reports buildkite/codewhale-ci.
CodeQL: alerts are still unreadable (403). GITHUB_CODEWHALE_SECURITY_PAT is not provisioned (4th week).
Fix the recurring flakes and drift jobs.
core::engine::approval::tests::execute_tools_gates_an_extension_tool_before_any_host_call: 11 failures, 2 of which passed on a same-SHA rerun.
The timing-based macOS persistence-backlog-budget: enqueue_elapsed_ns=33309879 exceeds 25000000 on mainb1313572; also flagged last week.
Marketplace connection :: catalog: First-party catalog drift: run python3 scripts/sync-marketplace.py, failing on main HEAD.
Housekeeping.
Lower the dead-code budget to 253 with --update (it is 5 under).
Triage the 97 unlabeled issues and the 12 external-author issues idle > 14 d (§6).
Required checks (ruleset 15930115): Lint, Test (windows-latest), npm wrapper smoke (ubuntu-latest). buildkite/codewhale-ci is not reported on any PR. No PR has unresolved review threads.
22 remote branches. Only release/0.9.12-prep-20260830 is older than 30 days with no open PR (last commit 2026-08-30, 36 d). There are no snapshot/* branches.
3. CI flake ranking (7 d)
GitHub Actions had 4,342 runs: 3,270 success, 196 failure, 279 cancelled, 98 action_required. (09-29 re-fetched in 6-hour windows to avoid the 1,000-run cap).
CI on main (push): 94 runs. 33 success, 5 failure, 56 cancelled. Last week: 72 / 31 / 53.
CI on PRs: 358 runs. 110 success, 63 failure, 168 cancelled, 16 action_required.
Buildkite codewhale-ci: unavailable. The API returns 404 No organization found (see action 3).
Failed jobs: Test win 41, ubuntu 35, mac 24; PR-closes link 32; Lint 25; Version drift 22; Web Lint 15; RC parity 9; Safety gate 8; npm smoke 7. (31 were compile errors.)
Flaky: a job failed, then passed on a rerun of the same SHA (20 job instances).
Same-SHA fail→pass
Test
Example
15
PR closes an issue :: link (body-edit rerun, not a test)
Other main failures: late_launch_permit_still_gets_the_full_work_budget (win), launch_card_pty::launch_resume_buttons… (mac), idle_managers_sharing_a_store_do_not_poll_it_continuously (win), Lint :: Check formatting (60ea7c22), the macOS budget timing check (×3), and CodeQL Setup at Checkout repository (b1313572).
Commands
gh api --paginate "repos/codewhale-hq/Codewhale/actions/runs?created=<day | 6h window>&per_page=100"# 4342 runs in window
gh api "repos/.../actions/runs/<id>/jobs?filter=all"# 665 runs (failed or run_attempt>1) -> 275 failed jobs
gh api "repos/.../actions/jobs/<id>/logs"| grep -E 'test (\S+) \.\.\. FAILED|---- (\S+) stdout ----'# 120 test-job logs# flaky = (sha, workflow, job) with a failed attempt n and a successful attempt > n
curl -H "Authorization: Bearer $BUILDKITE_READ_TOKEN" https://api.buildkite.com/v2/organizations/hunter-bown/pipelines/codewhale-ci
-> 404 No organization found (also for codewhale-hq, codewhale, hmbown)
4. Drift checks (fresh --depth 1 clone of main @ b059d4f0)
Check
Result
python3 scripts/check-dead-code-budget.py
253 suppressions (allow=171, expect=82), budget 258 (5 under). Down 7 from last week's 260.
Version state OK: workspace=0.10.1, npm=0.10.1, npm-binary=0.10.1, lockfile in sync.
Docs vs product. I parsed 106 registered commands and 96 aliases (non-test code) against backticked /cmd mentions in docs/ and web/.
In code, not documented (21):/advisor /anchor /branch /exit /fork /home /import-claude /logout /lsp /new /overview /pin /remote-env /save /structcopy /system /transcript /tree /verbose /voicecontrol /voicesend. /exit is in Help's curated COMMON_COMMANDS but appears in no doc.
Documented, not in code: 0 in user-facing docs. Last week's /search flag in docs/CONFIGURATION.md is a URL path, not a command. The rest are in design/archive docs or are the retired /set, which is kept only as migration guidance.
Settings:ConfigCategory::ALL has 7 categories (Appearance, ModelsProviders, Work, ToolsMcp, Trust, Motion, Advanced) and /setupSTEP_SPECS has 10 steps. No docs/ or web/ page enumerates either list (unchanged).
Read-only digest, window 2026-09-28 12:22Z → 2026-10-05 12:22Z. Previous: #6702. Repo now resolves to
codewhale-hq/Codewhale.Recommended actions (ranked by impact)
mainis probably red on the requiredTest (windows-latest)check (inferred). chore(deps): bump uuid from 1.26.0 to 1.27.0 #6826 (uuid-only, parent =mainHEADb059d4f0) fails 14 Windows tests (log) (extension_hostbun/native_mcpcluster,windows_native_lpac_bun_owner_boundary,runtime_store_convergence…).wave/0.10.1-next(0.10.1: contributor integration, human-wait lifecycle, and release qualification #6846, merged intomainate8036038) has failed the same set since 10-03 (log).main's own HEAD run was cancelled, so unconfirmed. This blocks all 13 Dependabot PRs.linux-arm64link error:undefined reference to 'renameat2'incodewhale_config::private_directory::move_socket_no_replaceon musl (log).action_required: fix(tools): flag fetch_url and web_run spill artifacts as retrievable evidence #6847–fix(web): bound the fetch SSRF pre-flight DNS lookup #6852, fix(tools): bound and reap pandoc conversions #6854–fix(snapshot): bound git subprocesses so a wedged git cannot stall the turn #6859, feat(automation): archive terminal runs on delete #6864 (13 from asto18089), plus feat(runtime-api): read what one tool call changed, from the snapshots around it #6817 (gaord), fix(tui): drop the commas the locale extraction put after ←, → and ⚠ #6844 (Lstarsky0) and feat(orcarouter): OAuth 2.0 + PKCE connect and a live chat catalog alongside the API key #6867 (hodeswildsmith455-boop). feat(plugins): support reviewed OAuth AI providers #6805, fix(search): honor the configured locale in Bing and DuckDuckGo scrapes #6860 and fix(client): bound non-streaming model requests with a retry-aware envelope #6863 areCONFLICTINGwith no runs; per AGENTS.md a maintainer resolves conflicts caused only bymainmoving.BUILDKITE_READ_TOKENauthenticates, butorganizations/hunter-bownreturns 404No organization foundand GraphQLviewer.organizationsis empty. The token appears to have lost org access, possibly with the move tocodewhale-hq. No PR reportsbuildkite/codewhale-ci.GITHUB_CODEWHALE_SECURITY_PATis not provisioned (4th week).core::engine::approval::tests::execute_tools_gates_an_extension_tool_before_any_host_call: 11 failures, 2 of which passed on a same-SHA rerun.persistence-backlog-budget:enqueue_elapsed_ns=33309879 exceeds 25000000onmainb1313572; also flagged last week.Marketplace connection :: catalog:First-party catalog drift: run python3 scripts/sync-marketplace.py, failing onmainHEAD.--update(it is 5 under).CONFLICTING) and branchrelease/0.9.12-prep-20260830.1. Open PRs: 34 open, 1 stale > 7 d
Required checks (ruleset 15930115):
Lint,Test (windows-latest),npm wrapper smoke (ubuntu-latest).buildkite/codewhale-ciis not reported on any PR. No PR has unresolved review threads.Commands
2. Stale branches: 1
22 remote branches. Only
release/0.9.12-prep-20260830is older than 30 days with no open PR (last commit 2026-08-30, 36 d). There are nosnapshot/*branches.3. CI flake ranking (7 d)
GitHub Actions had 4,342 runs: 3,270 success, 196 failure, 279 cancelled, 98 action_required. (09-29 re-fetched in 6-hour windows to avoid the 1,000-run cap).
CIonmain(push): 94 runs. 33 success, 5 failure, 56 cancelled. Last week: 72 / 31 / 53.CIon PRs: 358 runs. 110 success, 63 failure, 168 cancelled, 16 action_required.codewhale-ci: unavailable. The API returns 404No organization found(see action 3).Failed jobs: Test win 41, ubuntu 35, mac 24; PR-closes
link32; Lint 25; Version drift 22; Web Lint 15; RC parity 9; Safety gate 8; npm smoke 7. (31 were compile errors.)Flaky: a job failed, then passed on a rerun of the same SHA (20 job instances).
PR closes an issue :: link(body-edit rerun, not a test)Test (ubuntu)"Run tests" step, no test name parsed (fix/l1-config-secrets-crates)core::engine::approval::tests::execute_tools_gates_an_extension_tool_before_any_host_call(win)core::engine::tests::an_isolated_chat_engine_never_revokes_another_engines_extensionmcp::tests::streamable_http_reset_after_tool_call_post_is_not_replayed(mac),runtime_api::tests::restore_routes_refuse_an_active_turn_in_the_workspace,fleet::manager::tests::worker_exit_observed_after_the_deadline…,Pet conformance :: androidMost frequent failing tests (349 distinct; consistent per commit, so broken commits rather than flakes):
extension_host::tests::{memory_cap_stops_a_bun_host, bun_host_runs_the_dsh_plugin…, raw_agent_presets_*}+ 6extension_host::native_mcp::tests::*wave/0.10.1-nextand Dependabot PRs onmainHEADtools::subagent::tests::route_replacement::persona_receipt::actual_child_persona_selection…runtime_api::tests::runtime_store_convergence::real_owner_two_workspace_services…,extension_host::windows::tests::windows_native_lpac_bun_owner_boundaryconformance::events::golden_turn_events_match(incl.main96263571)extension_host::tests::memory_cap_stops_a_node_hostplugins::install::dsh::tests::*(≈20 tests)Other
mainfailures:late_launch_permit_still_gets_the_full_work_budget(win),launch_card_pty::launch_resume_buttons…(mac),idle_managers_sharing_a_store_do_not_poll_it_continuously(win),Lint :: Check formatting(60ea7c22), the macOS budget timing check (×3), andCodeQL SetupatCheckout repository(b1313572).Commands
4. Drift checks (fresh
--depth 1clone ofmain@b059d4f0)python3 scripts/check-dead-code-budget.py253 suppressions (allow=171, expect=82), budget 258 (5 under). Down 7 from last week's 260.Cargo.tomlworkspaceCHANGELOG.mdtop## [Unreleased], then## [0.10.1] - 2026-10-01✓npm/codewhale,npm/runtime-sdk,extensions/vscodenpm/deepseek-tui/webscripts/release/check-versions.sh --range-audit-advisoryVersion state OK: workspace=0.10.1, npm=0.10.1, npm-binary=0.10.1, lockfile in sync.Docs vs product. I parsed 106 registered commands and 96 aliases (non-test code) against backticked
/cmdmentions indocs/andweb/./advisor /anchor /branch /exit /fork /home /import-claude /logout /lsp /new /overview /pin /remote-env /save /structcopy /system /transcript /tree /verbose /voicecontrol /voicesend./exitis in Help's curatedCOMMON_COMMANDSbut appears in no doc./searchflag indocs/CONFIGURATION.mdis a URL path, not a command. The rest are in design/archive docs or are the retired/set, which is kept only as migration guidance.ConfigCategory::ALLhas 7 categories (Appearance, ModelsProviders, Work, ToolsMcp, Trust, Motion, Advanced) and/setupSTEP_SPECShas 10 steps. Nodocs/orweb/page enumerates either list (unchanged).5. Security posture
vulnerabilityAlerts(states:OPEN))/dependabot/alerts: 403)/code-scanning/alerts?state=open)Resource not accessible by integration;GITHUB_CODEWHALE_SECURITY_PATnot provisioned6. Open issues: 186 (+19 vs 167)
42 opened and 32 closed in the window.
Labels: unlabeled 97, enhancement 47, needs-triage 31, bug 13, rust 9, tui/documentation/cleanup 5 each, tools/reliability/bot-authored 3 each, responses-api/agent-ready 2 each, subagents/question/model-lab/external-memory 1 each.
No human comment in more than 14 days: 90 (measured by the latest non-bot comment, or creation date if none).
gitCLI reads with gix (gitoxide) #5618 (LmeSzinc), Request: zero-sandbox / --no-sandbox mode for local dev #4955 (eugenicum), 输出内容中的文件,能不能支持点击后打开预览,这样不用再去目录中找 #2342 (caeserchen), FreeBSD support (npm binary / pkg) #1097 (Gabriel-Degret), 未适配中文输入法 #2323 (cmdcorp6534), Continuous Self-Learning from Dialog (Automatic Skill Evolution) #5860 (Edouard-Legoupil), feat(tui): consider Ctrl+C clearing a non-empty composer before it arms exit (Esc/Ctrl+C split revisited) #6237, feat(goal): goal gates — independent verification of complete/blocked, needs_input/deferred/stalled statuses, post-verify stage, gate resilience (config-optional; default = current behavior) #6013 and feat(tui): usage & tool diagnostics — token accounting (per-component, per-model + cache hit rate, per-tool sinks, compaction cost) and tool-call error patterns #6011 (7jrxt42BxFZo4iAnN4CX).Commands