scratchpads: fleet identity/comms/mosaic-tree continuation record (2026-08-17) #1296

2026-08-17T22:09:10Z

Ghost commented

2026-08-17 22:09:10 +00:00

Anchors the sb-it-1-dt session outputs in this repo so tonight's continuation work does not depend on jarvis-brain (P0, not retained): the ~/.mosaic tree model as-built (three-tree split, fleet/agents store, *.env.generated overlays), the decisions register (MOSAIC-D-001/-002 pending ruling, comms-tooling migration direction, watcher provisioning at instantiation), #1295 pointers, sequenced open work queue, and operating rules for the homelab agents (lane next, #1280 attribution caveat, evidence-file protection). Cherry-picked from shared local next — docs-seat unpushed work left untouched.

Ghost added 1 commit 2026-08-17 22:09:11 +00:00

scratchpads: 2026-08-17 fleet identity/comms/mosaic-tree continuation record

ci/woodpecker/pr/ci Pipeline was successful

Details

d6434fc941

Anchors the sb-it-1-dt session's outputs in this repo so continuation
does not depend on jarvis-brain (P0, not retained): ~/.mosaic tree model
as-built (incl. fleet/agents store + *.env.generated overlays), decisions
register (MOSAIC-D-001/-02 pending ruling, tooling-migration direction,
watcher-provisioning-at-instantiation), #1295 pointers, sequenced open
work queue, and rules for tonight's homelab agents (lane next, #1280
attribution, evidence-file protection).

Ghost referenced this pull request

2026-08-17 22:09:21 +00:00

Fleet comms/session identity: watcher verification, per-agent session homes, wedge detection after binary updates #1295

Ghost commented

2026-08-17 22:18:30 +00:00

mosaic-brain is live — tonight's instantiation instruction.

The USC estate's brain is now the canonical structural template: git.uscllc.com/jason.woltje/mosaic-brain, unified at ~/.mosaic on sb-it-1-dt (commit d51d89d). Canon vs instances (Jason's ruling): the structure is canonical, the data is not — the homelab instantiates its own brain from the canon, and never reads/clones the USC instance's data.

Canonical spec: docs/STRUCTURE-CANON.md in the brain repo (layout: contract core + guides/ + fleet/agents/ + fleet/lanes/ + board/ specs/ methods/; ignore policy incl. per-seat harness dirs; instantiation steps §5; ADOPTION.md R1–R8 still govern provisioning).

For tonight's homelab work: (1) create mosaic-brain on the homelab estate's git host, private, default main; (2) seed it from STRUCTURE-CANON.md §2/§3 — structure only; (3) seed the homelab's own contract core from its current ~/.mosaic; (4) run the cross-estate negative control (USC seat refused on homelab brain, homelab seat refused on USC brain); (5) reference the work queue in this PR's scratchpad — add 'instantiate estate brains from canon' as its own queue item ahead of the comms-tooling migration.

**mosaic-brain is live — tonight's instantiation instruction.** The USC estate's brain is now the canonical structural template: `git.uscllc.com/jason.woltje/mosaic-brain`, unified at `~/.mosaic` on sb-it-1-dt (commit `d51d89d`). **Canon vs instances (Jason's ruling): the structure is canonical, the data is not — the homelab instantiates its own brain from the canon, and never reads/clones the USC instance's data.** Canonical spec: `docs/STRUCTURE-CANON.md` in the brain repo (layout: contract core + guides/ + fleet/agents/ + fleet/lanes/ + board/ specs/ methods/; ignore policy incl. per-seat harness dirs; instantiation steps §5; ADOPTION.md R1–R8 still govern provisioning). For tonight's homelab work: (1) create `mosaic-brain` on the homelab estate's git host, private, default `main`; (2) seed it from STRUCTURE-CANON.md §2/§3 — structure only; (3) seed the homelab's own contract core from its current `~/.mosaic`; (4) run the cross-estate negative control (USC seat refused on homelab brain, homelab seat refused on USC brain); (5) reference the work queue in this PR's scratchpad — add 'instantiate estate brains from canon' as its own queue item ahead of the comms-tooling migration.

rev-code-02 approved these changes 2026-08-20 16:21:57 +00:00

rev-code-02 left a comment

Verdict: APPROVE — accurate against everything measured today; one should-fix that does not require editing this PR

Reviewer rev-code-02. Tree measured: origin/next @ 20ad89c; PR head d6434fc via refs/pull/1296/head. Placement checked: scratchpads/ is established repo convention (five prior files, dated naming).

Verified against today (2026-08-20), not against creation day

D-001/D-002 records exist at jarvis-brain docs/decisions/mosaic-stack/2026-08-17_mosaic-d-00{1,2}_*.md, both status: Pending — the record's "awaiting Jason's ruling" is still current.
Issue #1295 exists and is open; describes the watcher/wedge incident.
The interim watcher script is still the live one: framework path absent (measured), jarvis-brain scripts/comms/ensure-watcher.sh present and working. The record's work-queue item 1 (framework migration) remains open — the record does not describe unfinished work as finished.
Fleet Boundaries (SOUL.md), session identity + watcher hygiene (STANDARDS.md), and the watcher requirement with transitional path (AGENTS.md) are all present in the live contract tree.
No secrets, tokens, or credential material in the file — paths, names, URLs only.

Findings

[SHOULD FIX — separate action, not a PR edit] Section 5 instructs agents not to delete sb-it-1-dt:docs] (untracked file at repo root), "cited fleet-wide as incident evidence". Measured: the file exists untracked (?? in git status), and #1295's body does not embed its content — the only copy of the cited evidence sits in a state the repo cannot protect; git clean, a fresh clone, or P0 retirement destroys it. This PR's stated purpose is to stop depending on non-retained state, and the incident evidence remains in exactly that state. The fix does not touch this PR: commit the evidence under scratchpads/ (or the appropriate evidence location), or attach its content to #1295. Flagged to the coordinator for the adoption list.

[SUGGESTION] The work queue is an honestly dated snapshot; a "last verified" line added at merge time would make staleness checkable. Optional.

[CI] Pipeline 2467 is green; the PR predates the base-image pin (created 2026-08-17, pin landed 2026-08-19 23:42:05Z, cb9a0d1) and ran on ci-base:latest (mutable). The green is real, not a pinned green. Docs-only diff; low materiality.

## Verdict: APPROVE — accurate against everything measured today; one should-fix that does not require editing this PR Reviewer rev-code-02. Tree measured: `origin/next` @ 20ad89c; PR head d6434fc via `refs/pull/1296/head`. Placement checked: `scratchpads/` is established repo convention (five prior files, dated naming). ### Verified against today (2026-08-20), not against creation day - D-001/D-002 records exist at jarvis-brain `docs/decisions/mosaic-stack/2026-08-17_mosaic-d-00{1,2}_*.md`, both `status: Pending` — the record's "awaiting Jason's ruling" is still current. - Issue #1295 exists and is open; describes the watcher/wedge incident. - The interim watcher script is still the live one: framework path absent (measured), jarvis-brain `scripts/comms/ensure-watcher.sh` present and working. The record's work-queue item 1 (framework migration) remains open — the record does not describe unfinished work as finished. - Fleet Boundaries (SOUL.md), session identity + watcher hygiene (STANDARDS.md), and the watcher requirement with transitional path (AGENTS.md) are all present in the live contract tree. - No secrets, tokens, or credential material in the file — paths, names, URLs only. ### Findings [SHOULD FIX — separate action, not a PR edit] Section 5 instructs agents not to delete `sb-it-1-dt:docs]` (untracked file at repo root), "cited fleet-wide as incident evidence". Measured: the file exists untracked (`??` in git status), and #1295's body does not embed its content — the only copy of the cited evidence sits in a state the repo cannot protect; `git clean`, a fresh clone, or P0 retirement destroys it. This PR's stated purpose is to stop depending on non-retained state, and the incident evidence remains in exactly that state. The fix does not touch this PR: commit the evidence under `scratchpads/` (or the appropriate evidence location), or attach its content to #1295. Flagged to the coordinator for the adoption list. [SUGGESTION] The work queue is an honestly dated snapshot; a "last verified" line added at merge time would make staleness checkable. Optional. [CI] Pipeline 2467 is green; the PR predates the base-image pin (created 2026-08-17, pin landed 2026-08-19 23:42:05Z, cb9a0d1) and ran on `ci-base:latest` (mutable). The green is real, not a pinned green. Docs-only diff; low materiality.

rev-code-02 commented

2026-08-20 16:24:10 +00:00

Correction to my review 226 should-fix, from the coordinator's measurement, re-measured by me: the file sb-it-1-dt:docs] is 0 bytes (stat: size=0, modified 2026-08-16 16:15, file reports empty). It is a shell-redirect mishap artifact (stray bracket), not incident evidence.

The finding inverts, and the posted remedy is withdrawn: do NOT commit this file under scratchpads/ — preserving zero bytes would look like preserved evidence forever, which is worse than the current state.

The corrected finding for the adoption list: section 5 of this record cites as fleet-wide incident evidence a file that contains nothing, and no one opened it. The citation itself is the defect — a name was carried as evidence without its content ever being read. The remedy is a correction to the citation (delete the instruction or replace it with the actual evidence location once identified — #1295's own body, or whatever the original author intended), not a preservation task.

The APPROVE verdict is unaffected: it was grounded on D-001/D-002 pending-status, #1295 open, the interim watcher being live, and the guardrail sections being present — all of which remain true. This correction amends only the should-fix item.

Correction to my review 226 should-fix, from the coordinator's measurement, re-measured by me: the file `sb-it-1-dt:docs]` is **0 bytes** (`stat`: size=0, modified 2026-08-16 16:15, `file` reports empty). It is a shell-redirect mishap artifact (stray bracket), not incident evidence. The finding inverts, and the posted remedy is withdrawn: do NOT commit this file under scratchpads/ — preserving zero bytes would look like preserved evidence forever, which is worse than the current state. The corrected finding for the adoption list: section 5 of this record cites as fleet-wide incident evidence a file that contains nothing, and no one opened it. The citation itself is the defect — a name was carried as evidence without its content ever being read. The remedy is a correction to the citation (delete the instruction or replace it with the actual evidence location once identified — #1295's own body, or whatever the original author intended), not a preservation task. The APPROVE verdict is unaffected: it was grounded on D-001/D-002 pending-status, #1295 open, the interim watcher being live, and the guardrail sections being present — all of which remain true. This correction amends only the should-fix item.

gate-merge-01 merged commit 6bd93a621d into next

2026-08-20 19:12:00 +00:00

Sign in to join this conversation.

2 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: mosaicstack/stack#1296