Skip to content

docs: record v1.11.0 and move fleet ops knowledge out of local notes - #29

Merged
mcarter-astronautdev merged 1 commit into
mainfrom
docs/v1110-handoff
Aug 2, 2026
Merged

docs: record v1.11.0 and move fleet ops knowledge out of local notes#29
mcarter-astronautdev merged 1 commit into
mainfrom
docs/v1110-handoff

Conversation

@mcarter-astronautdev

Copy link
Copy Markdown
Member

Post-wave documentation pass. No workflow or template changes — docs only.

CLAUDE.local.md is a capture buffer for to-dos, not a notebook, so everything durable that had accumulated in it is now tracked:

  • New docs/fleet-operations.md — the fleet inventory (why 21 and 18 are both right), why waves are direct pushes with [skip ci], the three wave traps, the three things fleet-pin-audit.sh structurally cannot see, and how to pilot a cross-repo reusable. README stays authoritative for the release order itself; this is the knowledge around it.
  • README — release-history sections for v1.9.0, v1.10.0 and v1.11.0 (the chain previously stopped at v1.8.0). v1.9.0's entry records that it never got its kit repin commit, because that is exactly the gap the audit reported clean through. The v1.11.0-specific wave warning is generalised into the durable rule and points at the new doc.
  • docs/reusable-conversion-scope.md — Phases 2 and 3 done, status COMPLETE, and the Phase 2 assertion resolved in place with the run evidence.

The driver-agents canonical-text follow-up moved to that repo's local to-dos, where the session that can act on it will see it. This repo's local file is now seven one-line to-dos, each pointing at a tracked doc for the detail.

…local notes

Everything durable that had accumulated in the gitignored CLAUDE.local.md is now
tracked. That file is a capture buffer for to-dos, not a notebook — anything
living only there is invisible to a fresh clone and to every other session.

New docs/fleet-operations.md — the operational knowledge around the release
order, which README stays authoritative for:
- the fleet inventory and why 21 and 18 are both correct numbers depending on
  the question (the 3-pair delta carries stub rails only)
- why waves are direct pushes with a skip-ci marker rather than 21 PRs: zero review
  runs, and zero theme deploys on the ~10 branches that have them
- the three traps — same basename in both halves of the diff, pin hunks that
  patch from a base the fleet was never on, and per-repo state that must survive
- what fleet-pin-audit.sh cannot see: files with no `uses:` line, file content,
  and drift in its own reference. All three bit during v1.9.0 -> v1.11.0.
- how to pilot a cross-repo reusable, including the two traps this pilot hit

README: release-history sections for v1.9.0, v1.10.0 and v1.11.0, and the chain
extended past v1.8.0. v1.9.0's entry records that it never got its kit repin
commit, since that is the gap the audit could not see. The v1.11.0-specific wave
warning in the release order is generalised into the durable rule (sed pin lines,
never git apply them) and points at fleet-operations for the rest.

reusable-conversion-scope.md: Phases 2 and 3 marked done, status flipped to
COMPLETE, and the Phase 2 assertion resolved in place with the evidence — the
pilot proved vars.BONSAI_URL resolves against the caller, so the per-repo tunnel
override survives. The two pilot-design corrections are recorded there and in
fleet-operations, because the claude.yml pilot would otherwise repeat them.

The driver-agents canonical-text follow-up moved to that repo's CLAUDE.local.md,
where the session that can act on it will see it.

(Commit message deliberately avoids the literal skip-ci token: GitHub scans the
whole message, not just the subject, so spelling it out here suppressed CI on
this very PR the first time round.)
@mcarter-astronautdev
mcarter-astronautdev merged commit 0e6d726 into main Aug 2, 2026
2 checks passed
@mcarter-astronautdev
mcarter-astronautdev deleted the docs/v1110-handoff branch August 2, 2026 13:57
mcarter-astronautdev added a commit that referenced this pull request Aug 2, 2026
…resses CI (#30)

GitHub scans the whole commit message, not just the subject, so a commit that
merely explains the wave technique matches its own skip marker and suppresses
every workflow on the resulting PR.

This already happened once: the commit that added fleet-operations.md described
the pattern in its body, and PR #29 came back with zero check runs. That failure
mode is nastier than a red check — absent checks look identical to checks that
have not started yet, so the PR reads as healthy and an admin merge would sail
through unvalidated.

Recorded next to the wave procedure, since that is the context in which someone
will next be tempted to write the token in prose.
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