devBenches/base-image: re-vendor openRepoTools at b1015f4 with all fifteen installables by brettheap · Pull Request #134 · opensoft/workBenches

brettheap · 2026-10-05T18:24:34Z

Lane: openxfactory-5 (openXfactory-5)

Brett Heap's word, given in session on 2026-09-30, verbatim: "file and fix both". This is the second of the two; it is ruled on the lane register against workBenches#132, and lane openXfactory-5 holds the claim.

What this does

Re-vendors openRepoTools at b1015f4672f9c107d939803c39846f8dbd3d78f3, openRepoTools main's head when this was written. It contains daed2095 (opensoft/openRepoTools#134, merged 2026-10-04) and opensoft/openRepoTools#97.

Installables: fifteen. Counted from the vendored shim's own INSTALLABLES at that sha, not from the issue text: openRepoTools park resume status lane lanes lane-handoff lane-rename lanes-edit.sh lane-start lane-end link-estates repos.tsv claude-current claude-restart-check. The thirteen already vendored, plus claude-current and claude-restart-check at the end. SKILLS (3) and COMMANDS (3) are unchanged, so --install now places twenty-nine artifacts: 15 bin files, 6 skill files, 6 command files, 2 hook entries.

Why it matters: ~/.local/bin lives on the container overlay, and at every start workbench-entrypoint runs estate-commands-start, which reruns --install from the copy baked into the image under /usr/local/share/openrepotools. Until the image carries this pin, a bench start or recreate puts the 2026-09-15 bytes back over a newer install.

The image rebuild after this lands is Brett's act, not part of this PR. Nothing here was built, pushed or deployed.

Procedure, as the repo has it

python3 devBenches/base-image/update-upstream.py apply --source openrepotools --at b1015f4672f9c107d939803c39846f8dbd3d78f3 --yes --add claude-current --add claude-restart-check

7 re-taken (openRepoTools, lanes, lanes-edit.sh, lane-start, lane-end, lane, lane-handoff), 2 added, 12 unchanged byte for byte. update-upstream.py check: 23 vendored copies match (21 openrepotools + 2 openreposhape). Independently, all 15 bin files, the 3 SKILL.md and the 3 command files are cmp-identical to git show b1015f4:<path> in a separate clone of openRepoTools.

Files changed (15)

Vendored copies: devBenches/base-image/files/openrepotools/ (claude-current and claude-restart-check added; openRepoTools, lanes, lanes-edit.sh, lane-start, lane-end, lane, lane-handoff re-taken).
Pin rows: devBenches/base-image/upstream-pin.yaml (commit and sha256 rows, written by apply).
scripts/setup-estate-commands.sh: TOOLS_FILES carries the fifteen, which is what check_shim_list compares with the shim's own INSTALLABLES (a shim that disagrees refuses with "Nothing was installed"); header and inline counts re-derived.
devBenches/base-image/Dockerfile: COPY lines and the chmod 0755 block for the two new files; comments re-counted.
devcontainer.test/test-setup-estate-commands.sh: see below.
README.md and devBenches/base-image/files/estate/estate-commands-start: name lists and counts only (comments).

Two test changes that are not a pure count bump

The shim now writes an install receipt (--install writes a receipt of the files it placed, and a retirement reads that before it reads a header substring openRepoTools#103, ~/.local/share/openRepoTools/installed.tsv) and rewrites it on every run, "already installed" files included. Scenario (r), "a second start writes nothing", therefore failed at the new pin. Its manifest now leaves the receipt out, and the receipt is checked separately: the same name, destination and digest rows after a second start, only the UTC stamp may move.
New scenario (aa) reads the vendored shim's own INSTALLABLES and asserts the Layer 1a Dockerfile has one exact COPY line and one chmod word per name. Nothing compared the Dockerfile's per-file list with the shim before, which is how the image kept baking thirteen. Mutation-tested on a throwaway copy: dropping the claude-current COPY, dropping claude-restart-check from the chmod block, dropping lane while keeping lane-* (prefix collision), and removing it from the script's TOOLS_FILES each fail the suite.

Gates

Run twice in a scrubbed environment (env -i, clean HOME), tree clean: at 0a43d6b (the re-vendor commit) and again at a7f970d (the same commit with main's 23719de, the archive PR-B of #124, merged in; no conflicts, none of its paths overlap). Both runs identical and green. Only .github/workflows/speckit-git-bash.yml has a path filter this diff matches (all 15 changed paths hit it); the other five workflows match none.

update-upstream.py check: 23 copies match, rc 0.
test-setup-estate-commands.sh: GREEN, 469 PASS, 0 FAIL.
test-setup-workspace-repo.sh, test-base-image-dockerfile.sh, test-claude-profile-name-guard-hook.sh (59 assertions), test-claude-profile-binary-selection.sh (122, with the workflow's env block), test-claude-profile-skill-install.sh (142), test-claude-profile-amendment-11.sh (467), bootstrap, feature, compat and ct-launcher suites: all rc 0.
Docker jobs on the workflow's pinned images, at both heads: git-2-34-1 (ubuntu:22.04@sha256:2edbbc5d...) and bash-3-2 (bash:3.2@sha256:241dd7ed...): both GREEN.
bash -n and shellcheck -S error clean on both changed scripts; every COPY source in the Dockerfile exists.
NOT run: devBenches/scripts/test-helper-safety.sh (its workflow's path filter does not match this diff, and it writes the real shared MCP registry, so it never runs on this host); docker build of the image (the rebuild is Brett's act).

Review

CI at a7f970d: CodeQL, Analyze (actions, javascript-typescript, python), and regression, git-2-34-1, bash-3-2 of OpenSpeckit Bash: all success. Copilot reviewed 0a43d6b and a7f970d ("Changes recommended"); its 5 inline findings are all about the pinned upstream bytes, not this PR's own lines, so none is fixed here (an edit to a vendored copy is drift and update-upstream.py check refuses it). Each is already tracked upstream and was answered on its thread and resolved:

claude-current:445, claude-current: fall back from a broken newest native executable openRepoTools#154: the native resolver drops the newest candidate when its version probe fails instead of trying older ones.
claude-restart-check:206, claude-restart-check: recognize processes from custom native directories openRepoTools#155: detection honors only the default native directory.
openRepoTools:355, Refresh installer artifact and receipt count contracts after Claude resolver installation openRepoTools#148: the upstream comment's artifact and receipt counts are stale (27/25; they are 29/27 with the two new commands).
lanes-edit.sh:11820, tree_id_for is not injective and the recorded checkout is never compared with git's, so two trees can share one sidecar and a replaced repository reads as ok openRepoTools#116: the sidecar's recorded checkout is never compared with git's.
lane-handoff:1265, The handoff's READY line counts polled trees as recorded, and a stale registration is suppressed whenever a sidecar names the same path openRepoTools#117: the READY line counts polled trees as recorded.

The first two are functional defects in the two commands this re-vendor newly installs; a later re-vendor takes their fixes.

Sibling search

Open workBenches PRs at the time: #114, #115, #116, #118 (#133, archive PR-B of #124, landed during this work and is merged in; it touches none of these paths). None touches the pin, files/openrepotools/, setup-estate-commands.sh, the Dockerfile's estate block or test-setup-estate-commands.sh. Two soft overlaps, neither a conflict: #114 edits README.md in a different paragraph (the pclaude one); #118 edits devcontainer.test/test-base-image-dockerfile.sh, whose stale "twenty-six" / "twelve" comments I therefore left alone.

Closes #132

🤖 Generated with Claude Code

…fteen installables, so a bench start stops reinstalling the 2026-09-15 copy The openrepotools source pin moves from a7d1257 (openRepoTools#135) to b1015f4672f9c107d939803c39846f8dbd3d78f3, openRepoTools main's head at the time, which contains daed2095 (#134, "Launch current Claude Code and report stale running sessions") and #97 (crash-consistent lane worktree recovery). The installable set changed, not just the bytes. The vendored shim's own INSTALLABLES is now fifteen names: the thirteen it had, plus `claude-current` and `claude-restart-check`, listed last (#119's two Claude Code launch commands). SKILLS (three) and COMMANDS (three) are unchanged, so `--install` places twenty-nine artifacts: fifteen bin files, six skill files, six command files, two hook entries. `update-upstream.py apply --source openrepotools --at b1015f46... --yes --add claude-current --add claude-restart-check` took the set: seven re-taken (openRepoTools, lanes, lanes-edit.sh, lane-start, lane-end, lane, lane-handoff), two added, twelve unchanged byte for byte. `update-upstream.py check`: 23 vendored copies match. Every one of the fifteen bin files, the three SKILL.md and the three command files is also `cmp`-identical to `git show b1015f4:<path>` in an independent clone of openRepoTools. Every place that lists or counts the set moves with it: - scripts/setup-estate-commands.sh: TOOLS_FILES carries the fifteen, which is what check_shim_list compares with the shim's own INSTALLABLES (a shim that grew past it refuses with "Nothing was installed"); the counts in its header and inline comments are re-derived (fifteen bin files, twenty-one pinned openrepotools paths, twenty-nine artifacts). - devBenches/base-image/Dockerfile: COPY lines and the chmod block for claude-current and claude-restart-check beside the thirteen; the comments are re-counted. - devcontainer.test/test-setup-estate-commands.sh: TOOLS_FILES carries the fifteen; the `15 of 15 placed` lines; the D4 read-only scenario also asserts the two new files were not placed. The shim now writes an install receipt (openRepoTools#103, ~/.local/share/openRepoTools/installed.tsv) and rewrites it on every run, so scenario (r)'s "a second start writes nothing" manifest leaves it out and the receipt is checked separately (same rows, only the UTC stamp may move). New scenario (aa) reads the shim's own INSTALLABLES and asserts the Layer 1a Dockerfile has one COPY line and one chmod word per name: nothing compared the two lists before, which is how the image kept baking thirteen after the shim grew. - README.md and devBenches/base-image/files/estate/estate-commands-start: the name lists and counts. The image rebuild afterwards is Brett Heap's act, not this commit's. Lane: openxfactory-5 (openXfactory-5) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

…otools-b1015f4 No conflicts; main's side touches none of the re-vendored paths. Lane: openxfactory-5 (openXfactory-5) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Copilot

Copilot review overview

🟡 Changes recommended

The vendored resolver has two known functional defects that require upstream fixes and re-vendoring.

Review effort: Balanced
Findings: 2 · 1

Open (3)

Native resolver drops to no result instead of trying older versions · New
Custom native directory detection suppresses version replacement notices · New
Stale artifact and receipt counts after adding resolver commands · New

What changed in this PR

Re-vendors openRepoTools at b1015f4, expanding the image and host installer to all fifteen commands.

Changes:

Adds claude-current and claude-restart-check.
Synchronizes installer, Dockerfile, pin, documentation, and artifact counts.
Adds receipt and Dockerfile inventory regression coverage.

File	Description
`README.md`	Documents fifteen installed commands.
`scripts/setup-estate-commands.sh`	Validates and installs all fifteen tools.
`devcontainer.test/test-setup-estate-commands.sh`	Tests receipts and Dockerfile inventory.
`devBenches/base-image/upstream-pin.yaml`	Pins openRepoTools at `b1015f4`.
`devBenches/base-image/Dockerfile`	Installs both new commands.
`devBenches/base-image/files/estate/estate-commands-start`	Updates image command count.
`devBenches/base-image/files/openrepotools/openRepoTools`	Updates installer and receipt handling.
`devBenches/base-image/files/openrepotools/claude-current`	Adds Claude executable resolution and updates.
`devBenches/base-image/files/openrepotools/claude-restart-check`	Adds stale-session detection.
`devBenches/base-image/files/openrepotools/lanes`	Adds closed-lane visibility and actions.
`devBenches/base-image/files/openrepotools/lanes-edit.sh`	Adds upstream lane lifecycle operations.
`devBenches/base-image/files/openrepotools/lane-start`	Adds handoff, resolver, and recovery behavior.
`devBenches/base-image/files/openrepotools/lane-end`	Adds dormant-lane and ownership handling.
`devBenches/base-image/files/openrepotools/lane`	Adds handoff and safer live attachment.
`devBenches/base-image/files/openrepotools/lane-handoff`	Adds crash-consistent lifecycle recording.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

Copilot

Copilot review overview

🟡 Changes recommended

The vendored revision contains unresolved resolver, restart detection, and lifecycle inventory correctness defects.

Review effort: Balanced
Findings: 1 · 3 · 1

Open (5)

Verify repository identity before marking checkout healthy · New
Report successful inventory writes separately from polled trees · New
Custom native directory detection suppresses version replacement notices
Native resolver drops to no result instead of trying older versions
Stale artifact and receipt counts after adding resolver commands

brettheap · 2026-10-05T20:04:15Z

Lane: openxfactory-5 (openXfactory-5)

LANDING — lane openXfactory-5, session bbe4d255-02db-4578-840b-09924442e544@Eagle, 2026-10-05T20:04:20Z, PR #134 into opensoft/workBenches main

sourcery-ai · 2026-10-05T20:04:33Z

Sorry @brettheap, your pull request is larger than the review limit of 150,000 diff characters

brettheap · 2026-10-05T20:04:41Z

Lane: openxfactory-5 (openXfactory-5)

LANDED — lane openXfactory-5, session bbe4d255-02db-4578-840b-09924442e544@Eagle, 2026-10-05T20:04:47Z, PR #134 into opensoft/workBenches main → merge 70e4aa8; devBenches/base-image: re-vendor openRepoTools at b1015f4 with all fifteen installables

Copilot AI balanced review requested due to automatic review settings October 5, 2026 18:24

Copilot started reviewing on behalf of brettheap October 5, 2026 18:25 View session

Merge origin/main (23719de, PR-B of #124) into chore/revendor-openrep…

a7f970d

…otools-b1015f4 No conflicts; main's side touches none of the re-vendored paths. Lane: openxfactory-5 (openXfactory-5) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Copilot AI reviewed Oct 5, 2026

View reviewed changes

Comment thread devBenches/base-image/files/openrepotools/claude-current

Comment thread devBenches/base-image/files/openrepotools/claude-restart-check

Comment thread devBenches/base-image/files/openrepotools/openRepoTools

Copilot AI balanced review requested due to automatic review settings October 5, 2026 18:28

Copilot started reviewing on behalf of brettheap October 5, 2026 18:28 View session

Copilot AI reviewed Oct 5, 2026

View reviewed changes

Comment thread devBenches/base-image/files/openrepotools/lanes-edit.sh

Comment thread devBenches/base-image/files/openrepotools/lane-handoff

brettheap marked this pull request as ready for review October 5, 2026 20:04

brettheap merged commit 70e4aa8 into main Oct 5, 2026
12 checks passed

brettheap mentioned this pull request Oct 6, 2026

Land adopt-triad-feature-amendments: the adopted Triad feature-amendment workflow and its OpenSpec change #135

Merged

Navigation Menu

Repository navigation

devBenches/base-image: re-vendor openRepoTools at b1015f4 with all fifteen installables - #134

devBenches/base-image: re-vendor openRepoTools at b1015f4 with all fifteen installables#134
brettheap merged 2 commits into
mainfrom
chore/revendor-openrepotools-b1015f4

brettheap commented Oct 5, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

brettheap commented Oct 5, 2026

Uh oh!

Uh oh!

sourcery-ai Bot commented Oct 5, 2026

Uh oh!

brettheap commented Oct 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

brettheap commented Oct 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this does

Procedure, as the repo has it

Files changed (15)

Two test changes that are not a pure count bump

Gates

Review

Sibling search

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

🟡 Changes recommended

Uh oh!

Uh oh!

Uh oh!

brettheap commented Oct 5, 2026

Uh oh!

Uh oh!

sourcery-ai Bot commented Oct 5, 2026

Uh oh!

brettheap commented Oct 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

brettheap commented Oct 5, 2026 •

edited

Loading