DEN-3945: consolidate all branches into main (2026-09-12 sweep)

[ORESoftware]

Consolidation of all branches into main — DEN-3945 fleet sweep (2026-09-12)

Linear: DEN-3945. Every local branch, worktree and uncommitted change under ~/codes was published without force; this PR merges every branch not already in main with merge commits only, conflicts resolved semantically per ORESoftware/my-ai/AGENTS.md. Durable sweep record: ORESoftware/den-3945-fleet-sweep.

Semantically resolved and merged

• chore/repository-local-worktrees-20260905 — Merge 8a05194. Fleet repository-local worktree policy that never landed here (no PR was ever opened: head query and commits/7dcc158/pulls both empty at rc 0) and is already on main in sibling repos, so it is the outstanding copy, not a competing design. This repo tracks AGENTS.md and agents.md as two DIFFERENT documents (AGENTS.md ends with the ore-primary-branch-policy block, agents.md with persistence-authority), and the branch appends the same 7-line section to both; the two appended sections are byte-identical across the files. AGENTS.md applied cleanly because its base blob on the branch (b6ef21b) is the blob on current main. agents.md CONFLICTED and was resolved as a UNION, not by picking a side: both sides appended a new section right after , main adding Git and history policy (31 lines on rebase/stash/reset) and the branch adding Repository-local Git worktrees (7 lines). Topically disjoint and complementary, so both are kept, main first and the branch section last, matching where the same commit placed it in AGENTS.md. Result is a pure 7-line addition to main version of each file (14 lines total, zero deletions). No .gitignore change needed for internal consistency: the policy demands tmp, temp, tmp/worktrees and temp/worktrees be ignored and main already ignores all four.
• chatgpt/ores-compose-infra-fleet-20260914 — Merge 5e02078, conflict-free: 610 insertions across 4 files, every one absent from origin/main, so nothing is overwritten. Adds the fail-closed ores-compose infra fleet rollout plus its format-candidate and lock-candidate workflows. Never landed and never proposed: head query and commits/f3a2421/pulls both empty at rc 0. Checked against standing fleet decisions before merging: the branch publishes no container image name so DECISIONS.md 17c does not apply; the new tooling is bash, not Python, per the standing preference; and the workflows reach GH_TOKEN through the secrets context rather than embedding a credential.

Not merged — already landed / superseded (evidence)

• DEN-666-ores-trace-org-admission-20260913 — Landed. Tip 5038542 is itself a commit on origin/main, merged through PR DEN-666: reusable static ores-trace admission #95 (merge commit f0e7efd); became a strict ancestor of the sweep once origin/main was merged in.
• docs/merge-evidence-pr-template-20260914 — Landed. Tip 2d56311 is literally a commit on origin/main (PR docs: define exact-head CI evidence policy #97, merged 2026-09-14T20:20:31Z); ancestor of the sweep after the origin/main merge.
• automation/2026-08-19-preserve-client-hardening — c2-identical-tree-on-main; head PR fix: preserve reviewed client hardening branches #47 merged 2026-08-20T05:01:09Z into main.
• automation/run-now-20260902 — NET-ZERO branch. c1-every-changed-path-landed, and git diff --stat against its merge-base with the sweep head is completely EMPTY: its 3 commits add a one-shot fleet task dispatcher (360e45c, 6c35931) and then remove it again (2c47e44 ops: remove one-shot fleet task dispatcher). No PR was ever opened (commits/2c47e44/pulls empty, rc 0). There is nothing to carry; merging it would only reintroduce its 3 ops commits with no content change.
• docs/exact-head-ci-evidence-20260914 — c2-identical-tree-on-main; head PR docs: define exact-head CI evidence policy #97 merged 2026-09-14T20:20:31Z (the squashed successor 2d56311 is on main).
• fix/public-reusable-cli-portability-20260915 — c2-identical-tree-on-main; head PR ci: add public reusable CLI portability workflow #104 merged 2026-09-15T16:24:17Z (squashed onto main as 8771248).
• governance/contract-boundary-r2-20260914 — c2-identical-tree-on-main; head PR docs: codify fleet contract dependency boundary #98 merged 2026-09-14T21:09:08Z (squashed onto main as 5b852f0).
• governance/immutable-release-provenance-r2-20260914 — c2-identical-tree-on-main; head PR docs: require immutable release provenance #99 merged 2026-09-14T21:29:34Z (squashed onto main as b0bf4cb).
• recovered/heads/main-297ce587fe67-20260822 — Content landed and was then superseded on main. classify reports c1-every-changed-path-landed; the one path the branch alone carries is .github/pull_request_template.md (blob 70d0cad6, its docs: streamline pull request guidance commit 1869d29 at 2026-08-09T19:23:49). git log --find-object shows that exact blob on origin/main at c230d13 docs: reconcile pull request governance template, 2026-08-09T19:27:32, four minutes later, which took the streamlined text and then reconciled it into .github/PULL_REQUEST_TEMPLATE.md while DELETING the lowercase .github/pull_request_template.md. c230d13 is an ancestor of origin/main. Merging the branch would resurrect the deleted lowercase file next to the uppercase one, creating a second competing PR template and a new case collision inside .github/, and would revert main by ~4800 lines of later work.

Not merged — the branch has its own open PR

• chatgpt/zed-tip-fleet-sync-20260914 — ci: synchronize Zed dependency tips across 100+ repos #96
• automation/harden-scheduled-jobs-20260820 — fix(automation): harden nightly dispatch, pins, and audit roles #48

Not merged — needs a human/product decision

• none

Local verification (exact head 25cd34ce08d1)

ALL 73 origin branches were tested for containment, not just the 2 the claim listed; the claim predated 6 new commits on origin/main and 9 branches created after the sweep was built. Clone made self-contained: repack -a -d, alternates removed, fsck --connectivity-only clean (only dangling objects, no missing/broken links); refs/local/* (4 refs incl. refs/local/main) all 0 commits ahead of origin/main, so no unpushed local work was stranded. open-pr-heads rc=0 listing exactly 2 open PRs, both skipped. FIRST ACTION was merging origin/main (dec94c9..8771248) into the sweep, which is what reduced 14 uncontained branches to 2 genuinely unlanded ones. CASE COLLISION: this repo tracks both AGENTS.md and agents.md (and three PR-template case variants: .github/PULL_REQUEST_TEMPLATE.md, PULL_REQUEST_TEMPLATE.md, pull_request_template.md), so EVERY commit was made with git merge-tree --write-tree plus commit-tree/update-ref and a temporary GIT_INDEX_FILE; the working tree was never used to merge and nothing was ever staged from it. Proof the caution was needed: extracting the merged tree with git archive onto this macOS filesystem yields only agents.md and only pull_request_template.md, the other case variants being silently clobbered. Both collision blobs were verified by rev-parse in every resulting tree. Gates from .github/workflows run on the merged head with origin/main as the baseline: actionlint 4 findings on the merged head and the SAME 4 on origin/main (3x SC2195/SC2221/SC2222 in reusable-cli-portability.yml, 1 typing note in tjsv-admission-selftest.yml), i.e. pre-existing, and the 3 newly added workflows are actionlint-clean; scratch-ignore-policy gate PASSES on the merged head (scratch_ignore_check.rs compiled with -D warnings, 10/10 unit tests pass, checker reports 4 directory probes ignored and 4 same-named file probes visible, argument-refusal probe exits 2 as the workflow requires) - note it was built with local rustc 1.95.0 rather than the pinned 1.85.1; shellcheck -S warning on the new tools/ores_compose_infra_fleet.sh reports 2 SC2155 declare-and-assign warnings and bash -n is clean, and no workflow in the repo runs shellcheck over tools/, so nothing is gated by them; git diff --check origin/main HEAD clean. Not run: ores-lint (its outcome is unchanged by construction - the merge touches no .js/.mjs/.ts/.rs/.dart/.gleam file) and the fleet-wide audit workflows, which need live org credentials and a runner. Net vs origin/main: 6 files, +624 lines, 0 deletions. Pushed after every single commit; remote sweep head verified by ls-remote after each push and equals the local head. Disk 55 GB free; all extraction and build output I created was deleted. NOTE for whoever next opens this clone: its working tree and index are still at the pre-merge checkout, so git status is noisy - that is cosmetic, every commit is on GitHub.

Contracts (TJSV)

Run locally on the repo peer-authority fixture pair test-fixtures/tjsv-admission/main.tsp vs authored.schema.json (the pair the reusable-tjsv-admission selftest feeds), TJSV eac6163 - note the workflow pins the action at 4a5d049. Result on the MERGED HEAD: STOPPED_FOR_EVALUATION, 1 structural finding, generated-authored-semantic-mismatch for Zed.Governance.AdmissionFixture at #/unevaluatedProperties, differential lane 30 probes 30 agreements 0 divergences 0 refusals. PRE-EXISTING, not caused by this sweep: the identical run on origin/main returns exactly the same status and the same single finding, and both authorities are byte-identical between main and the merged head (main.tsp e8a1834e, authored.schema.json f5b74eef), so the merge cannot change the verdict. Neither authority was regenerated from the other. The selftest workflow is path-filtered to reusable-tjsv-admission.yml, tjsv-admission-selftest.yml and test-fixtures/tjsv-admission/**, none of which this sweep touches, so it would not even fire. No camelCase/snake_case question arose: the fixture fields are single words (id, count).

Merge policy

Draft until verified on the exact head. Where this org has no GitHub Actions minutes, zero-step jobs are admission refusals, not passes.

🤖 Generated with Claude Code

[linear-code]

DEN-3945