Skip to content

0.0.9 milestone run: coordinator tracking #2474

Open

Description

@rickylabs

Tracking issue for the 0.0.9 milestone run.

A long-running milestone coordinator drives every open 0.0.9 issue to closed or moved-with-a-reason, except the design-system epic (#2412 and its epic:design-system leaves), which the Fresh UI design-system lane owns.

  • Plan: "fix what dogfooding broke". Critical path H01 → A1/A2/A3 → B1 → B2 → terminal canary. If late, cut F1 and E2 first, never A or B.
  • Each issue ships as its own leaf PR with a cross-family review at the exact head and green CI.
  • Eric cuts every canary and the stable release. The coordinator posts readiness here and never tags or publishes.
  • Status comments land here at least twice a day. Questions for Eric go here, each with a recommendation.

Activity

  1. added this to the 0.0.9 milestone on Oct 30, 2026
  2. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    /swarm
    repo: rickylabs/netscript
    harness: claude
    tier: complex
    role: coordinator
    profile: milestone-coordinator
    model: claude-opus-5-5
    effort: xhigh
    max-tokens: 300m
    timeout: 168h
    approval: bypass

    use harness. You are the 0.0.9 milestone coordinator, run 2026-10-30 mc009. This issue is your tracking issue. Read BRIEF-0.0.9-coordinator.md in the scratch netscript-release folder of your home first; it is your full brief and wins over anything here.

    Scope: every open 0.0.9 issue except label epic:design-system (the Fresh UI design-system lane owns those; never touch them).

    You coordinate and never implement. Dispatch each issue as its own leaf, merge only on READY plus a cross-family MERGE at the exact head plus fully green CI, keep your context lean (state file plus a short STATUS.md, compact when needed), and post status here at least twice a day. Eric cuts every canary and the stable release: post readiness here and keep the other lanes moving. Run until every non-design 0.0.9 issue is closed or moved with a reason and "stable cut ready" is posted.

    Golden rules in order: 1 the idiomatic NetScript way (in-repo MCP find_guidance first); 2 SOLID; 3 performance; 4 the Doctrine.

    Rules: never print secrets, tokens, operator paths, host names, ports or session ids in public comments; never restart ai-agents, herdr, orchid, the forwarders or Tailscale; never kill by pattern; never force-push; no milestone or label changes without Eric's yes. Everything else here is approved: never wait for an answer.

  3. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: launch refused. Reason: source-grant-missing. A comment launches only with a valid Cockpit grant for that exact comment. A final missing-grant refusal means its admission window has closed or no grant endpoint is available. Nothing was launched. Post a new /swarm comment to try again.

  4. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    /swarm
    repo: rickylabs/netscript
    harness: claude
    tier: complex
    role: coordinator
    profile: milestone-coordinator
    model: claude-opus-5-5
    effort: xhigh
    max-tokens: 300m
    timeout: 168h
    approval: bypass

    use harness. You are the 0.0.9 milestone coordinator, run 2026-10-30 mc009b (relaunch after an orchid restart). This issue is your tracking issue. Read BRIEF-0.0.9-coordinator.md in the scratch netscript-release folder of your home first; it is your full brief and wins over anything here.

    Scope: every open 0.0.9 issue except label epic:design-system (the Fresh UI design-system lane owns those; never touch them).

    You coordinate and never implement. Dispatch each issue as its own leaf, merge only on READY plus a cross-family MERGE at the exact head plus fully green CI, keep your context lean (state file plus a short STATUS.md, compact when needed), and post status here at least twice a day. Eric cuts every canary and the stable release: post readiness here and keep the other lanes moving. Run until every non-design 0.0.9 issue is closed or moved with a reason and "stable cut ready" is posted.

    Golden rules in order: 1 the idiomatic NetScript way (in-repo MCP find_guidance first); 2 SOLID; 3 performance; 4 the Doctrine.

    Rules: never print secrets, tokens, operator paths, host names, ports or session ids in public comments; never restart ai-agents, herdr, orchid, the forwarders or Tailscale; never kill by pattern; never force-push; no milestone or label changes without Eric's yes. Everything else here is approved: never wait for an answer.

  5. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: launch refused. Reason: source-grant-missing. A comment launches only with a valid Cockpit grant for that exact comment. A final missing-grant refusal means its admission window has closed or no grant endpoint is available. Nothing was launched. Post a new /swarm comment to try again.

  6. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    /swarm
    repo: rickylabs/netscript
    harness: claude
    tier: complex
    role: coordinator
    profile: milestone-coordinator
    model: claude-opus-5-5
    effort: xhigh
    max-tokens: 300m
    timeout: 168h
    approval: bypass

    use harness. You are the 0.0.9 milestone coordinator, run 2026-10-30 mc009c (relaunch after the grant socket fix). This issue is your tracking issue. Read BRIEF-0.0.9-coordinator.md in the scratch netscript-release folder of your home first; it is your full brief and wins over anything here.

    Scope: every open 0.0.9 issue except label epic:design-system (the Fresh UI design-system lane owns those; never touch them).

    You coordinate and never implement. Dispatch each issue as its own leaf, merge only on READY plus a cross-family MERGE at the exact head plus fully green CI, keep your context lean (state file plus a short STATUS.md, compact when needed), and post status here at least twice a day. Eric cuts every canary and the stable release: post readiness here and keep the other lanes moving. Run until every non-design 0.0.9 issue is closed or moved with a reason and "stable cut ready" is posted.

    Golden rules in order: 1 the idiomatic NetScript way (in-repo MCP find_guidance first); 2 SOLID; 3 performance; 4 the Doctrine.

    Rules: never print secrets, tokens, operator paths, host names, ports or session ids in public comments; never restart ai-agents, herdr, orchid, the forwarders or Tailscale; never kill by pattern; never force-push; no milestone or label changes without Eric's yes. Everything else here is approved: never wait for an answer.

  7. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: automatic launch abandoned (registration_failed). The agent was not confirmed ready, or disappeared after supervision. No automatic respawn will occur for this trigger, including after dispatcher restart. Inspect the failed launch before requesting a new dispatch with a new /swarm comment.

  8. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: automatic launch abandoned (registration_failed). The agent was not confirmed ready, or disappeared after supervision. No automatic respawn will occur for this trigger, including after dispatcher restart. Inspect the failed launch before requesting a new dispatch with a new /swarm comment.

  9. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: matrix launch inconclusive. Reason: launch-inconclusive. Field: launch. Launch effects could have occurred; inspect the durable fence and the agent before any new dispatch. Launch outcome requires inspection; this notice does not authorize another attempt. Cause: spawn.workspace-create.

  10. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: automatic launch abandoned (registration_failed). The agent was not confirmed ready, or disappeared after supervision. No automatic respawn will occur for this trigger, including after dispatcher restart. Inspect the failed launch before requesting a new dispatch with a new /swarm comment.

  11. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: matrix launch inconclusive. Reason: launch-inconclusive. Field: launch. Launch effects could have occurred; inspect the durable fence and the agent before any new dispatch. Launch outcome requires inspection; this notice does not authorize another attempt. Cause: spawn.workspace-create.

  12. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: matrix launch inconclusive. Reason: launch-inconclusive. Field: launch. Launch effects could have occurred; inspect the durable fence and the agent before any new dispatch. Launch outcome requires inspection; this notice does not authorize another attempt. Cause: spawn.workspace-create.

  13. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    /swarm
    repo: rickylabs/netscript
    harness: claude
    tier: complex
    role: coordinator
    profile: milestone-coordinator
    model: claude-opus-5-5
    effort: xhigh
    max-tokens: 300m
    timeout: 168h
    approval: bypass

    use harness. You are the 0.0.9 milestone coordinator, run 2026-10-30 mc009d (relaunch after herdr restore). This issue is your tracking issue. Read BRIEF-0.0.9-coordinator.md in the scratch netscript-release folder of your home first; it is your full brief and wins over anything here.

    Scope: every open 0.0.9 issue except label epic:design-system (the Fresh UI design-system lane owns those; never touch them).

    You coordinate and never implement. Dispatch each issue as its own leaf, merge only on READY plus a cross-family MERGE at the exact head plus fully green CI, keep your context lean (state file plus a short STATUS.md, compact when needed), and post status here at least twice a day. Eric cuts every canary and the stable release: post readiness here and keep the other lanes moving. Run until every non-design 0.0.9 issue is closed or moved with a reason and "stable cut ready" is posted.

    Golden rules in order: 1 the idiomatic NetScript way (in-repo MCP find_guidance first); 2 SOLID; 3 performance; 4 the Doctrine.

    Rules: never print secrets, tokens, operator paths, host names, ports or session ids in public comments; never restart ai-agents, herdr, orchid, the forwarders or Tailscale; never kill by pattern; never force-push; no milestone or label changes without Eric's yes. Everything else here is approved: never wait for an answer.

  14. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    divybot: started. claude is working on this issue on n5-agents with claude-opus-5-5.

  15. rickylabs commented on Oct 30, 2026

    @rickylabs
    OwnerAuthor

    0.0.9 coordinator status: Step 0 done, wave 1 dispatched (2026-10-30 22:40 UTC)

    Run record: #2475, a draft that stays open for the whole run. The milestone validator is green at baseline main 30c800958.

    Scope: 53 issues in four lanes, seven waves. That is every open 0.0.9 issue minus the 19 epic:design-system issues (Fresh UI lane) and this tracking issue.

    • fixes: 18
    • features: 8
    • internals: 21
    • docs: 6

    Deadlines (approved; 18:00 Paris): canary .1 on 11-02, .2 on 11-06, .3 on 11-08, freeze and .4 on 11-11, stable on 11-14. I post readiness here as soon as a canary's contents are merged and green.

    Dispatched (wave 1, Codex sol):

    Group Issue Leaf
    H01 #2438 OpenTelemetry fix and lock-resolved temp-config fixtures
    A1 #2165 scheduled-event ids stable across restarts
    A1 #2187 file-watch dedup, in parallel with #2165 under split file ownership
    A3 #2144 retries for in-process jobs, then a job DLQ
    D1 #2396 adopts PRs #2421–#2423 (audit, rebase, fix); no duplicate

    The opposite-family PLAN-EVAL is running in parallel with wave 1, as you ordered dispatch now. Its findings apply from wave 2.

    Next: wave 2 (#2395, #2400, #2151, #2161, #2185) as builder slots free (cap 5 builders, ≤ 6 launches/h).

    Questions for Eric. Each has a recommendation. The lanes keep moving meanwhile.

    1. feat(ai): one-shot audio chat completion for unauthenticated ASR endpoints #2465 placement. Recommend Backlog / Triage: it is a new AI capability, not a dogfooding fix. Needs your yes for the milestone move.
    2. "Code quality" red on main since feat(docs): render file trees and GitHub alerts with shared components #2469. The unsafe-cast is at docs/site/_plugins/markdown-render-fixture.ts:48; PR CI is unaffected because it runs the diff-scoped job. Recommend filing a one-leaf 0.0.9 issue (internals), which I would dispatch in wave 2. Needs your yes for the new issue in 0.0.9.
    3. PRs fix(docs): qualify complete examples and identities on Windows #2424 and fix(cli): install complete Aspire workflows deterministically #2433 are in 0.0.9 with no linked issue. Recommend moving both to Backlog / Triage unless you want them in this release.
    4. Failed-canary policy. Recommend: keep independent work moving, stop the affected lanes, block stable.
    5. DS1 stable gate. Recommend the S-slices gate stable and the review follow-ups move at freeze if not done.
    6. Auto-merge. If orchid auto-merges green leaf PRs on this repo, please turn it off for the 0.0.9 run. Otherwise a leaf could merge before the cross-family MERGE verdict at its head. Leaf PRs stay drafts until I merge them.
  16. 68 remaining items

  17. rickylabs commented on Nov 1, 2026

    @rickylabs
    OwnerAuthor

    0.0.9 coordinator: paused (17:38 UTC, under the Claude-budget rule: all Claude lanes pause at 80% weekly until the Friday reset). Cockpit keeps #2488 moving, reviews and merges on green. Eric cuts the canaries while this lane is paused.

    State

    canary.2 gate (pre-approved by Eric, like every canary until stable)

    1. feat(fresh-ui): complete contracts and core component gaps #2488 (S3 phase 2) merges. It's in flight now.
    2. Main push CI is green at the exact sha: ci, Code quality, Generated carrier safety net and Fresh UI package quality.
    3. Dispatched controls pass on that sha: gh workflow run e2e-cli.yml --ref main and gh workflow run command-postgres.yml --ref main both end green, and main has not moved.
    4. Cut: gh workflow run release-canary.yml --ref main -f target-version=0.0.9. Confirm the release/canary-pair status and e2e-cli-prod are green.
    5. Qualify: send the exact version to Cockpit for the atelier-cockpit smoke, with results posted here.

    Next actions, in order

    1. canary.2, per the gate above.
    2. The Fresh UI lane closes its gate list: [cli] Scaffold redesign: Figma archetype pages, capability-aware with designed empty states #2570, [cli] Opt-in showcase scaffold: Northwind commerce demonstrating services, workers, sagas, streams, AI and MCP #2571, [docs] /create and component pages follow-ups: payload budget, find_guidance routing, remaining pickers #2569, [design-system S7] Theme builder and component docs #2420 S7 phase 1, [design-system S3] Component contract, data-attribute variants, and missing components #2416 S3 phase 3, the small follow-ups, the bookkeeping closes and [design-system] Add faithful platform themes: macOS 27, Apple Classic and Fluent 2 Web #2443. At freeze, S6b and the residual items move to Frontend Contribution.
    3. The terminal canary on the final content, through the same gate, then the atelier-cockpit smoke on it.
    4. Post "stable cut ready" with the evidence: the canary pair, e2e-cli-prod, the cockpit smoke and exact-main CI. Stable is on Eric's go.
    5. Close 0.0.9 milestone run: coordinator tracking #2474 through the coordinator branch PR.

    The run record is on draft PR #2475: state, STATUS, worklog, cut trace and receipts.

  18. rickylabs commented on Nov 1, 2026

    @rickylabs
    OwnerAuthor

    Cockpit: PR 2488 (S3 phase 2) MERGED 18:54Z as a182c37 after DeepSeek v4.1 Flash review MERGE (b7663d3; delta to bdfe31c = main merge + 2 small doc/test fixes, CI green). The canary.2 gate is clear for Eric to cut. Also merged today: 2582 (#2578, 4a2dae3) and 2583 (#2579, 5117549).

  19. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    0.0.9-canary.2: published, production E2E red (08:25 UTC)

  20. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    Eric (in session, 08:36 UTC): main reopens now. Independent work keeps merging, and #2593 is the only gate for 0.0.9-canary.3. canary.3 re-qualifies whatever main holds when it is cut.

  21. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    0.0.9-canary.3 published green (09:50 UTC)

  22. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    Consumer smoke: atelier-cockpit on 0.0.9-canary.3

    • Version: every @netscript/* pin moved to exactly 0.0.9-canary.3 (--minimum-dependency-age=0). Consumer at atelier-cockpit e5a7108.
    • OS: Linux 6.18. Deno 2.9.7, TypeScript 6.0.3.
    • DB: in-memory PGlite 0.4.3 (embedded PostgreSQL 17.5). The 20 opt-in external-PostgreSQL cases were skipped.
    • Canary: release-canary run 41420257713 and e2e-cli-prod 41421575852, both green.
    • Gates: check 1, lint 0, fmt:check 0, test 1, test:production-permissions 1.

    Result: no upstream regression. All 10 distinct failure causes are consumer-side adaptations to intended 0.0.9 behaviour:

    1. Worker health callbacks may now return Promise<ChildHealthSnapshot> (consumer type update).
    2. The consumer's own test:production-permissions task is missing.
    3. The consumer pins LogTape 2.4.0 directly while the framework resolves 2.4.2.
    4. Saga replay keys are now versioned and hashed (consumer assertions).
    5. A synthetic watcher fixture must stay open until stop.
    6. Busy retry now preserves Retry-After (consumer tests assumed it was lost).
    7. Prisma projection v2 nests business state (state.state.phase).
    8. Announcement redrive dedupes per cron occurrence (tests must advance the clock).
    9. An undecodable foreign cookie no longer invalidates the valid application cookie.
    10. An authenticated unmatched route now returns 404 instead of 403.

    Each one passed in a local adaptation probe, with application logic unchanged. The consumer adopts 0.0.9 in its own repo; nothing here blocks the canary.

  23. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    Stable process (Eric, in session, 11:57 UTC):

    1. The terminal canary is cut on Fresh UI's final 0.0.9 content, under the usual gate and Eric's standing canary approval.
    2. Then I post "stable cut ready" with the evidence.
    3. Stable publishes only after Eric's explicit go at that point.

    The atelier-cockpit smoke on 0.0.9-canary.3 (above) shows no upstream regression. Its consumer adaptations are tracked in atelier-cockpit#979.

  24. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    0.0.9 scope update (owner decision, 2026-11-02): the #2571 Northwind showcase ships in 0.0.9, and the stable release waits for it to merge. The fresh-ui gate items before the terminal canary are PRs #2559, #2573, #2574, #2592 (#2591) and the #2571 PR. Deferred to 0.0.10: the contract promotions in code (S3 phase 3), the remaining styles in code (S6b), Code Connect and the Figma follow-up passes.

  25. rickylabs commented on Nov 2, 2026

    @rickylabs
    OwnerAuthor

    Stable cut ready: 0.0.9 (terminal canary 0.0.9-canary.4)

    The terminal sha is main d0f53933f5c1ddd7953fd86eb203227236217084. Every gate item for the 0.0.9 stable release is merged:

    Each PR had a cross-family review verdict of MERGE at its exact head, plus a leak scan.

    Evidence on d0f53933:

    • main ci 41536813111, Code quality 41536813106, carrier safety net 41536813085, Fresh UI package quality 41536813109 and Pages 41536813187: all success.
    • e2e-cli 41536868868: success.
    • release-canary 41539714417 published 0.0.9-canary.4: success.
    • e2e-cli-prod against the published packages, 41540843051: success.
    • Consumer smoke on canary.3: no upstream regression (see the comment above). The consumer adaptations are tracked downstream.

    Deferred to 0.0.10: the S3 phase 3 promotions, S6b, Code Connect, the S4 sign-off, the Figma passes, and the #2597 showcase follow-ups.

    Waiting for the owner's explicit go before publishing stable.

  26. rickylabs commented on Nov 3, 2026

    @rickylabs
    OwnerAuthor

    NetScript 0.0.9 stable is released: v0.0.9, marked Latest and not a prerelease.

    Cut PR: #2598. Merge: 9aac52aa0.

    Publish 41582229212 and e2e-cli-prod 41582759632 are green. All 35 @netscript packages from the canary.4 baseline list 0.0.9 on JSR.

  27. rickylabs commented on Nov 7, 2026

    @rickylabs
    OwnerAuthor

    ⏱️ divybot: operator timeout of 168h0m0s exceeded; native work stopped; seat and process absence observed. Retry requires a new /swarm comment.

  28. rickylabs commented on Nov 11, 2026

    @rickylabs
    OwnerAuthor

    Consumer-qualified SDK defect filed as #2610, blocking Garage source87's unchanged native GET regression. NetScript0.0.9/oRPC1.16.1 default GET dedupe replaces logical context with group.context {}, then SDK fetch overrides init.cache/signal from that empty context. Exact pinned source and current-main source confirm it; public options expose no supported remedy. Reproducer/acceptance are in #2610 (actual cache+abort/deadline/concurrent caller behavior), with no private operational payloads. Please triage through the existing NetScript maintainer/SDK route; no new Garage builder or consumer workaround is requested. A supported framework fix/coherent published cohort is needed before consumer requalification; Garage's red gate is retained. This is not Cockpit/trace/maintenance dependency and does not authorize a release or bypass.

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions