You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
bump-leg.py --dry-run, step 1 of advancing a leg, is refused bump-leg-on-tracking-branch on every fresh project, before the old -> new line the human is to read #477
(lane openRepoShape-3, 2026-10-12T21:54Z) Found by lane 3's v1.0.0 acceptance pass: an assistant following AGENTS.md literally on an ordinary scaffolded project, Linux, standard at 36538fd. Lane 3 claims it.
What is wrong. "Advancing a leg" step 1 is python3 scripts/bump-leg.py --root <path> --leg spec --to <40 hex> --dry-run, and "show the human the printed old -> new line" (AGENTS.md:323-328). On a fresh project, every one of which stands on its tracking branch after setup, that dry run is refused:
REFUSED bump-leg-on-tracking-branch ...
It exits 2 before the old -> new line the human is meant to read. The refusal exists to keep the COMMIT off the tracking branch, and a dry run commits nothing. AGENTS.md never says to make a branch first, yet two lines later it pushes <branch>. README.md:974-977 and its worked example at :502-505 skip the same step. The refusal's own remedy (git switch -c …) works, so the assistant recovers, but only by improvising a step the procedure omits.
tests/test_bump_leg.py:1006 covers the refusal without --dry-run only.
Proposed fix.
Let --dry-run run on the tracking branch: it writes nothing. Have it say, after the old -> new line and the trailers, that the real run needs a branch, and name the git switch -c command.
The real run keeps its refusal byte for byte.
In AGENTS.md (within the cap) and README, make creating the branch an explicit step before the real run, so <branch> in the push line has an origin.
Tests: a dry run on the tracking branch prints the line and the branch hint and exits 0; the real run there is still refused; the dry run on a branch is unchanged.
Why v1.0.0. Step 1 of a core procedure fails as written on every new project. Proposed P2.
CLAIMED — lane openRepoShape-3, session 20b4ae33-927d-4b5d-a856-e7e467ea11b2@Eagle, 2026-10-12T22:02:51Z, for #477
Logged in lanes/log/openRepoShape-3.md at 21ab2506cad68c51e873d13dd03a67f35eb75a91.
The three reads (lane-collision-protocol Rule 1):
1. existing `CLAIMED —` comments on opensoft/openRepoShape#477:
none
2. `gh pr list --repo opensoft/openRepoShape --state all --search "#477"`, kept where the row or its body names `#477`:
none
3. `git ls-remote --heads git@github.com:opensoft/openRepoShape` branches naming `477` as a whole token:
none
Activity
brettheap commentedon Oct 12, 2026
CLAIMED — lane openRepoShape-3, session 20b4ae33-927d-4b5d-a856-e7e467ea11b2@Eagle, 2026-10-12T22:02:51Z, for #477
Logged in
lanes/log/openRepoShape-3.mdat21ab2506cad68c51e873d13dd03a67f35eb75a91.The three reads (lane-collision-protocol Rule 1):
bump-leg.py: the `git switch -c` a `--dry-run` on the tracking branch…
bump-leg.py: `--dry-run` runs on the project's tracking branch and en…
Merge origin/main into fix/docs-validate-pins-gh-fallback (#411)
Merge origin/main (5a25387: #484 and four others) into docs/v1-accuracy
Merge origin/main (48585e6: #465 and three others) into docs/v1-accuracy
Merge origin/main (48585e6) into fix/doctor-mid-merge-not-compliant
Merge origin/main (48585e6: #465) into fix/setup-no-terminal-route
Merge origin/main (aa4e1f5) into fix/scaffold-reuse-seeded-legs
bump-leg.py and update-shape.py quote the paste lines that named a ro…
bump-leg.py and setup-project.py: the `git switch -c` line says under…