docs: canonicalize starros-stack-spec — San auditor pass v1.1 #1

Merged
j7 merged 1 commit from spec/starros-stack-canonicalization into main 2026-07-09 01:54:49 -04:00
Collaborator

Summary

Replaces the 2201-byte stub at specs/starros-stack-spec.md with a 26995-byte canonical spec (15 sections) derived from the live state of Mononoke (verified 2026-07-09) and J7's editorial pass.

What's in it

  • §2 Architectural Layers — table of all live/planned components with state badges ([WORKING] / [PARTIAL] / [PLANNED] / [BROKEN]); reflects J7 directive (GBrain in, Honcho out, Karakeep live @ keep.starrwulfe.xyz, CryptPad to be added, external coder workers = 4 CLIs + Claude Code)
  • §3 Core Architecture — corrects prior misdiagnosis (Ashitaka is unbound but is NOT a duplicate-multiplex hazard); clarifies multiplex is running on :8642 with multiplex_profiles: false
  • §7 UX / Workspace Model — Mission Control marked as J7-paused (not flake-input broken)
  • §8 Known uncertainty — Hermes Workspace reframed as intentionally killed by J7 (recurrent crashes), to revisit later
  • §9 Source-of-Truth Map — 22-row table tying every concern to a canonical file + owner
  • §10 Component Ownership Map — adds GBrain (TBD owner), Karakeep (live), external coder workers row; strikes Honcho
  • §11 End-to-End Task Flow — adds Matrix status-room rule (one room per project, wolves post status) per J7 directive
  • §12 Rollout Phases — renumbered to 13 phases; P11 = GBrain substrate spec (NEW per J7); P6 = add Claude Code; Hermes Workspace moved to P2 deferred
  • §13 Top 10 Risks — Mnemosyne dashboard now #1; Hermes Workspace and Honcho marked intentionally de-scoped
  • §14 Recommendations — add GBrain spec; diagnose Ashitaka port 8643

Verified facts (audit, 2026-07-09)

  • Multiplex PID 490006 holds :8642; PID 20888 (Ashitaka) is alive but unbound (not a duplicate-multiplex hazard)
  • Per-wolf ports 8644-8652 are bound by alphabetical systemd instantiation; :8643 is the intended slot for Ashitaka per wolfpack.nix defaults
  • Karakeep is live at keep.starrwulfe.xyz (HTTP 200, PocketID OIDC button present)
  • Mnemosyne dashboard :8765 is down (not "running on some port")
  • 4/5 external coder CLIs installed at /run/current-system/sw/bin/: opencode, kilocode, gemini, copilot. claude is not present
  • 11 wolf SOUL.md files exist at /home/san/.hermes/profiles/<wolf>/SOUL.md
  • Agent cards: only san.yaml + ashitaka.yaml written out of 11

Edit history

  • v1.0 (2201 bytes, J7 original, commit 908aae4a)
  • v1.1 (26995 bytes, this PR, commit 21dee6d) — full canonicalization

Sources of truth used

  • ADR 0001 (decisions/0001-repo-is-source-of-truth.md), ADR 0002
  • docs/playbook.md, docs/workflow.md
  • kanban/{board-schema,backlog}.md
  • ~/.hermes/wolf-souls/<wolf>/ and ~/.hermes/profiles/<wolf>/SOUL.md
  • /nas/hdd/agent-shared/agent-cards/wolves/{san,ashitaka}.yaml
  • /nas/hdd/agent-shared/wiki/projects/starrOS-to-megapack-transfer.md
  • /etc/nixos/nix/services/wolfpack.nix
  • ~/.hermes/skills/devops/hermes-multi-profile-architecture/references/wolfpack-dashboard-data-contract.md
  • Two prior San audit reports: ~/.hermes/cron/output/2026-07-07_mononoke_superapp_architecture_dossier.md and …_wolfpack_audit_round2.md

Not in this PR (held for follow-up)

  • GBrain substrate spec (Phase P11)
  • 9 missing Agent Cards (Phase P9)
  • Claude Code install (Phase P6)
  • Mnemosyne dashboard restart (Phase P4)
  • Hermes Workspace revival (Phase P2 deferred)

Test plan

  • J7 visual review of all 15 sections
  • Confirm Mnemosyne scoping language matches J7's intent (shared DB with author_id/channel_id filtering)
  • Confirm GBrain placement (in scope, undefined, Phase P11)
  • Confirm Karakeep owner swap (was San-consumer → now Mokku-taxonomy)
  • Jigo review (per ADR 0001)
  • Ashitaka review (multi-domain decomposition)

— San

## Summary Replaces the 2201-byte stub at `specs/starros-stack-spec.md` with a 26995-byte canonical spec (15 sections) derived from the live state of Mononoke (verified 2026-07-09) and J7's editorial pass. ## What's in it - **§2 Architectural Layers** — table of all live/planned components with state badges (`[WORKING]` / `[PARTIAL]` / `[PLANNED]` / `[BROKEN]`); reflects J7 directive (GBrain in, Honcho out, Karakeep live @ keep.starrwulfe.xyz, CryptPad to be added, external coder workers = 4 CLIs + Claude Code) - **§3 Core Architecture** — corrects prior misdiagnosis (Ashitaka is unbound but is NOT a duplicate-multiplex hazard); clarifies multiplex is running on `:8642` with `multiplex_profiles: false` - **§7 UX / Workspace Model** — Mission Control marked as J7-paused (not flake-input broken) - **§8 Known uncertainty** — Hermes Workspace reframed as intentionally killed by J7 (recurrent crashes), to revisit later - **§9 Source-of-Truth Map** — 22-row table tying every concern to a canonical file + owner - **§10 Component Ownership Map** — adds GBrain (TBD owner), Karakeep (live), external coder workers row; strikes Honcho - **§11 End-to-End Task Flow** — adds Matrix status-room rule (one room per project, wolves post status) per J7 directive - **§12 Rollout Phases** — renumbered to 13 phases; **P11 = GBrain substrate spec** (NEW per J7); **P6 = add Claude Code**; Hermes Workspace moved to P2 deferred - **§13 Top 10 Risks** — Mnemosyne dashboard now #1; Hermes Workspace and Honcho marked intentionally de-scoped - **§14 Recommendations** — add GBrain spec; diagnose Ashitaka port 8643 ## Verified facts (audit, 2026-07-09) - Multiplex PID 490006 holds `:8642`; PID 20888 (Ashitaka) is alive but unbound (not a duplicate-multiplex hazard) - Per-wolf ports `8644-8652` are bound by alphabetical systemd instantiation; `:8643` is the intended slot for Ashitaka per `wolfpack.nix` defaults - Karakeep is live at `keep.starrwulfe.xyz` (HTTP 200, PocketID OIDC button present) - Mnemosyne dashboard `:8765` is **down** (not "running on some port") - 4/5 external coder CLIs installed at `/run/current-system/sw/bin/`: `opencode`, `kilocode`, `gemini`, `copilot`. `claude` is not present - 11 wolf SOUL.md files exist at `/home/san/.hermes/profiles/<wolf>/SOUL.md` - Agent cards: only `san.yaml` + `ashitaka.yaml` written out of 11 ## Edit history - v1.0 (2201 bytes, J7 original, commit `908aae4a`) - v1.1 (26995 bytes, this PR, commit `21dee6d`) — full canonicalization ## Sources of truth used - ADR 0001 (`decisions/0001-repo-is-source-of-truth.md`), ADR 0002 - `docs/playbook.md`, `docs/workflow.md` - `kanban/{board-schema,backlog}.md` - `~/.hermes/wolf-souls/<wolf>/` and `~/.hermes/profiles/<wolf>/SOUL.md` - `/nas/hdd/agent-shared/agent-cards/wolves/{san,ashitaka}.yaml` - `/nas/hdd/agent-shared/wiki/projects/starrOS-to-megapack-transfer.md` - `/etc/nixos/nix/services/wolfpack.nix` - `~/.hermes/skills/devops/hermes-multi-profile-architecture/references/wolfpack-dashboard-data-contract.md` - Two prior San audit reports: `~/.hermes/cron/output/2026-07-07_mononoke_superapp_architecture_dossier.md` and `…_wolfpack_audit_round2.md` ## Not in this PR (held for follow-up) - GBrain substrate spec (Phase P11) - 9 missing Agent Cards (Phase P9) - Claude Code install (Phase P6) - Mnemosyne dashboard restart (Phase P4) - Hermes Workspace revival (Phase P2 deferred) ## Test plan - [ ] J7 visual review of all 15 sections - [ ] Confirm Mnemosyne scoping language matches J7's intent (shared DB with `author_id`/`channel_id` filtering) - [ ] Confirm GBrain placement (in scope, undefined, Phase P11) - [ ] Confirm Karakeep owner swap (was San-consumer → now Mokku-taxonomy) - [ ] Jigo review (per ADR 0001) - [ ] Ashitaka review (multi-domain decomposition) — San
Replaces the 2201-byte stub with a 26995-byte canonical spec derived from
the live state of Mononoke (verified 2026-07-09) and J7's editorial pass:

- §2 Architectural Layers — table of all live/planned components with
  state badges; reflects J7 directive (GBrain in, Honcho out, Karakeep
  live @ keep.starrwulfe.xyz, CryptPad to be added, external coder
  workers = 4 CLIs + Claude Code)
- §3 Core Architecture — corrects prior misdiagnosis (Ashitaka is
  unbound but is NOT a duplicate-multiplex hazard); clarifies multiplex
  is running on :8642 with multiplex_profiles=false
- §7 UX — Mission Control marked as J7-paused (not flake-input broken)
- §8 Known uncertainty — Hermes Workspace reframed as intentionally
  killed by J7 (recurrent crashes), to revisit later
- §10 Component Ownership Map — adds GBrain (TBD), Karakeep (live),
  external coder workers row; strikes Honcho
- §11 E2E Task Flow — adds Matrix status-room rule (one room per
  project, wolves post status) per J7 directive
- §12 Rollout Phases — renumbered to 13 phases; P11 = GBrain substrate
  spec; P6 = add Claude Code; Hermes Workspace moved to P2 deferred
- §13 Top 10 Risks — Mnemosyne dashboard now #1; Hermers Workspace and
  Honcho marked intentionally de-scoped
- §14 Recommendations — add GBrain spec; diagnose Ashitaka port 8643

Edit history block in preamble documents the v1.0 → v1.1 transition.

Verified against live ground truth (commit sha cdaaa0263 on main);
no other files in the repo touched; AGENTS.md/README.md in the
parent workspace dir /nas/hdd/agent-shared/WORKSPACES/starros-v1/
remain in place as scratchpad.

Refs: ADR 0001 (repo = source of truth), wolfpack.nix (canonical
runtime), kanban/board-schema.md, /nas/hdd/agent-shared/wiki/projects/
project-megapack.md, /nas/hdd/agent-shared/wiki/projects/starrOS-to-
megapack-transfer.md, ~/.hermes/cron/output/2026-07-07_wolfpack_audit_
round2.md.

San <san@mononoke.local>
j7 merged commit 99b5b49dbe into main 2026-07-09 01:54:49 -04:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
j7/StarrOS!1
No description provided.