Skip to content

docs: fleet is 23 pairs now, and why a kit-only file needs no wave - #32

Open
mcarter-astronautdev wants to merge 1 commit into
mainfrom
docs/fleet-count-23
Open

docs: fleet is 23 pairs now, and why a kit-only file needs no wave#32
mcarter-astronautdev wants to merge 1 commit into
mainfrom
docs/fleet-count-23

Conversation

@mcarter-astronautdev

Copy link
Copy Markdown
Member

Merging DriverDigital/driver-agents#6 and DriverDigital/driver-agents-app#2 moved the fleet 21 → 23, which makes several present-tense counts in these docs wrong. Numbers here are re-verified against a live audit run, not incremented by hand:

$ tools/fleet-pin-audit.sh
reference: templates/github/ pins all at v1.11.0 (90f0d066)
latest: v1.11.0 (90f0d066) — pins by SHA:
 110 90f0d066
content: 131 match templates/, 0 drifted     exit 0
before after
repo@branch pairs 21 23
pin rows 108 110
content matches 127 131
drift 0 0

Changes:

  • Repin-wave targets 21 → 23.
  • Stub-rails-only set 3 → 5driver-agents@main and driver-agents-app@main join Team-Laird@develop, The-Gathery@develop and driver-bonsai-mcp@main. Full-kit targets stay 18: the two new repos took the partial install, so neither carries claude.yml or bonsai-status-sync.yml.
  • Branch-protection survey now covers 23 pairs. Both new repos require one approving review, so the count lacking a human-approver rule is unchanged at 10 — still the 8 Palmers branches plus studio-sulzer@main and Team-Laird@develop.
  • The partial install is no longer described as "proposed". It is what those two repos run.

The part worth reading

This landed a new kit file (templates/github/lint.yml) with no tag and no wave, which is a departure from every prior kit addition, so the reasoning is now written down rather than left to be re-derived.

It was correct here because lint.yml changes no reusable, repins no stub, and carries no uses: DriverDigital/workflows@<sha> of its own — so nothing already deployed had to move, and the audit read clean throughout.

The non-obvious part is which way the obligation runs: cutting the tag is what creates the wave. The moment the latest tag is not 90f0d066, the audit's reference check fires against all six stubs, templates/ must be repinned, and the fleet must be re-copied — 23 branches — for a release whose reusables nobody touched. So a kit-only file should ride along with the next release that actually changes a reusable.

v1.7.0 deliberately did the opposite ("stubs repinned regardless, purely so the latest-tag comparison stays meaningful"), and that was right, because that wave was carrying kit content to the fleet. This one wasn't.

…wave

Merging the two agent-repo installs moved the fleet 21 -> 23, which makes
several present-tense counts wrong. Re-verified against a live audit run rather
than incremented by hand: 23 pairs, 110 pin rows at 90f0d06, 131 files matching
templates/, zero drift.

- Repin-wave targets 21 -> 23.
- The stub-rails-only set 3 -> 5: driver-agents@main and driver-agents-app@main
  join Team-Laird, The-Gathery and driver-bonsai-mcp. Full-kit targets stay 18 —
  the two new repos take the partial install, so neither carries claude.yml or
  bonsai-status-sync.yml.
- The branch-protection survey is now 23 pairs; both new repos require one
  approving review, so the count lacking a human-approver rule stays 10.
- The partial install is no longer "proposed" — it is what those two repos run.

Also writes down the question this raised, since the answer is not obvious and
the cost of guessing wrong is a 21-branch wave: a kit-only addition needs
neither a tag nor a wave when it changes no reusable, repins no stub, and
carries no uses: pin of its own. The obligation runs the other way — cutting the
tag is what creates the wave, because once the latest tag is not 90f0d06 the
reference check fires against all six stubs and templates/ has to be repinned
and re-copied everywhere.
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.

1 participant