adopt-project.py execute: each leg repository is asked about only as it is made, so a code leg whose name is taken leaves the spec leg created, on GitHub as under --local-remote-dir, and a --local-remote-dir that is a file or cannot be written ends in a traceback after check said plan ok #798

Goal: execute asks about both leg repositories, and about where they go, before it makes either, and check asks the same. So a taken name, or a --local-remote-dir that cannot hold them, is refused with nothing made.

Filed by lane openRepoShape-1 from #457's lines L1-72 and L1-73 (#457, comment) on Brett Heap's word of 2026-10-17T10:21:32Z, under lane 3's v1.1.1 rule (#473, issuecomment-6216697223). Each line was measured again on main bf1f099b with a stub gh.

Milestone: v1.1.1. On GitHub, an organisation that already holds <P>-code gets a stray private <P>-spec. Removing it needs the delete_repo scope. #188's N4 gave this case a notice; nothing prevents it.

Lines

• L1-73, local run: the code leg's bare repository already exists. With <lrd>/Atlas-code.git there, check --local-remote-dir <lrd> says plan ok. execute then gives REFUSED creating the code leg repository. leg-remote-exists: …/Atlas-code.git already exists, exit 2, after Atlas-spec.git was made.
• L1-73, GitHub run: the code leg's name is taken. This was simulated with a gh stub that logs every call and answers the code create with GitHub's Name already exists on this account (createRepository).
  • The only gh calls were gh repo create acme/Atlas-spec --private …, then gh repo create acme/Atlas-code --private ….
  • The run ended NOTHING has been rolled back: \https://github.com/acme/Atlas-spec.git\` exists`.
• L1-72: an unusable --local-remote-dir. check --local-remote-dir says plan ok in all three of these cases:
  • a file at the path: execute exits 1 with a FileExistsError traceback at _create_leg_remotes;
  • a path under a parent this user cannot write in: a PermissionError traceback;
  • an existing directory this user cannot write in: REFUSED a git or gh command failed … git init -q --bare … Permission denied, exit 2.

Direction

• Before _create_leg_remotes makes anything, ask both legs.
  • Under --local-remote-dir: whether either bare path exists (os.path.lexists), and whether the directory is, or can be made, a directory this user can write in, probed as _unwritable_work_dir probes a --work-dir. Refuse every finding at once, with NO_LEG_YET.
  • Against GitHub: gh repo view <org>/<P>-spec and <org>/<P>-code, taking only gh's not-found answer as "absent", as scaffold-project.py and family.py: both runs ask whether each folder they make can be made and take only gh's not-found answer as "absent", so --dry-run refuses a file where a folder goes, or a gh repo view that fails otherwise, by name as the real run does, and a refused init leaves no empty family folder (#609) #622 does for the scaffold. A taken name is refused before any repo create. A view that fails otherwise is refused by name, as the scaffold refuses it.
• check reports the same findings under the same flags, in the same order, as adopt-project.py check takes no --local-remote-dir and resolves a relative submodule url from GitHub's remotes, while execute --local-remote-dir resolves it from the bare repositories it creates, so under that test path check refuses an adoption execute verifies and passes one execute refuses #417's "one verdict" asks. The PR says what check asks of GitHub, and whether it reads the network there.

Tests: extend tests/test_adopt_execute_refusals.py and tests/test_adopt_late_failure.py.

• With <P>-code.git already there, the run is refused and no <P>-spec.git is made.
• A file, a path under an unwritable parent, and an unwritable directory as --local-remote-dir each give exit 2 with no traceback.
• When a gh stub's repo view finds <P>-code, there is no repo create call.
• check names each case.

Overlaps: #417's check --local-remote-dir, which landed as #718.

Lane: openreposhape-1 (openRepoShape-1)

🤖 Generated with Claude Code