[brettheap]

Lane: openreposhape-2 (openRepoShape-2)
Claim: #600 (comment)

What changed

adopt-project.py plan writes its follow-ups once, and MOUNT_REMEDIATION and plan-leg-path-collides both tell a person to change legs.spec_path or legs.code_path in the plan afterwards; execute then mounts the legs at the edited paths. The read-across follow-up still named the plan-time values, so after that edit check's output, the split commit's message and the pull request's body (the same text) described a layout the project no longer had:

follow-up: code/src/app/main.py names `contracts/`, which moves to the spec leg: it must read across the assembly root — `../spec/contracts/` relative to `code/` — or take the path from the environment

over legs the plan mounted at elsewhere/c and elsewhere/s (measured on main fbbbc8d by the new tests, all four surfaces: plan's file, check, the split commit and the body gh pr create was given). It now reads:

follow-up: src/app/main.py in the code leg names `contracts/`, which moves to the spec leg: it must read across the assembly root — from the code leg, where the plan's `legs.code_path` mounts it, to `contracts/` in the spec leg, where `legs.spec_path` mounts it — or take the path from the environment

• L2-18 (cross_leg_references, follow_ups_for). The line says where the file is (in the code leg, by its path in the source) and names the two keys, as adopt-project.py: the Makefile block execute appends and the cross-leg follow-up name each leg where the plan mounts it, so --spec-path legs/spec exports a CONTRACTS_DIR that exists and a nested --code-path gets the right ../ path (#522) #572's Makefile line names legs.spec_path and adopt-project.py: the tracking branch defaults to the source's default branch, topics are set before the pull request, and NEXT names scripts/bootstrap.py when the shape's Makefile went aside (#479, #480, #481) #493-F3's names tracking_branch. It is true whatever the keys hold. cross_leg_references and follow_ups_for no longer take spec_path and code_path, so no value of either key can reach a line they write.
• The path from the code leg to the spec leg's contracts/ (rv600's F1, added in the second round below). The key-only line leaves it to the two mounts printed beside it, and adopt-project.py: a non-default --spec-path or --code-path is ignored by the adopted Makefile's CONTRACTS_DIR and by the cross-leg follow-up path #522's auA-13 showed a person gets it wrong for nested legs, and CONTRACTS_DIR covers contracts/ for the root Makefile alone. So the two sentences that already print the CURRENT mounts end with it, computed from the mounts they print: check's mount the legs at X/ and Y/ and the split message's mounted back here at X/ and Y/, which is also the pull request's body. spec_leg_from_code_leg, adopt-project.py: the Makefile block execute appends and the cross-leg follow-up name each leg where the plan mounts it, so --spec-path legs/spec exports a CONTRACTS_DIR that exists and a nested --code-path gets the right ../ path (#522) #572's helper that the first round had deleted, is back with its 11 rows in tests/test_adopt_contracts_dir.py and feeds contracts_from_code_leg. The stored line is unchanged.
• L2-19 (contracts/path-classification.yaml). The spec-governance rule's reason: said ../spec/contracts from code/, false for a nested plan and copied into every plan entry the rule decides (and into a placement plan from shape-doctor.py). It now says the code reads across the assembly-root mount, which is what the legs are mounted for. The guard that holds it path-free now covers every rule's reason: and question: and the default: block's (rv600's F2, second round).

Rounds

• First round (08b6b61, on main fbbbc8d): L2-18 and L2-19, as above.
• Second round (rv600's review of 08b6b61; head d64252a): git merge --no-ff origin/main at 8f453d0 (a1a6973, clean, tree 29ed2a06 as merge-tree predicted; it brings scaffold-project.py and adopt-project.py: a pin records digest_source: local-clone and never the --pin-source path, so no committed pin file holds a local directory, and a --pin-source nothing reads is pin-source-unused (#540) #570, install.sh: git-filter-repo with the uv already there where there is no pipx, never pip install --user, and two readinesses at the end, adoption: then done: (#595) #598, shape-doctor.py: the members row says make siblings places the working clones from a holder in a folder named after its repository, and that from anywhere else it clones nothing and prints the mv, as it does since #578 (#605) #624 and family.py: bump --to refuses a commit only the member's mount holds, reachable from no branch of the member's remote, as bump-leg.py refuses its twin, so a holder is never pinned to a commit nobody else can fetch (#613) #625, none with a hunk in common with this PR), then one fix commit:

  • F1. The follow-up stays key-only. The two sentences that already print the CURRENT mounts end with the path from the code leg to the spec leg's contracts/, computed from the mounts they print. check, for --spec-path legs/spec --code-path impl/code:

    ONE split commit on branch adopt/three-repo-shape of Northwind: `git rm -r` those paths, mount the legs at legs/spec/ and impl/code/, so from the code leg the spec leg's `contracts/` is `../../legs/spec/contracts/`
    

    and the split message, which is also the pull request's body (../spec/contracts/ for the defaults):

    ... mounted back here at legs/spec/ and impl/code/, so from the code leg the spec leg's `contracts/` is `../../legs/spec/contracts/`.
    

    After a hand edit of either key the path is the edited mounts' (legs/spec and code: ../legs/spec/contracts/).

  • F2. The LEG_PATH guard covers every rule's reason: and question: and the default: block's. rv600's mutant R4 now fails it.

  • F3. The two counts in Tests below name their sets.

  • F4 (a v1.1.0 release-notes sentence: re-run plan for a plan written before v1.1.0 whose legs.spec_path or legs.code_path was edited) is lane 3's, and not in this PR. For such a plan the old value-form line and the new sentence can both print, and disagree.

Why the follow-up names the keys and does not render from the plan (L2-18)

The issue offers two directions. Four surfaces carry the list: the plan file, check, the split commit and the pull request body. All read one stored text (_plan_follow_ups). A line that names the keys is true on every one by construction, and for any edit of either key, including one made between plan and check. Rendering at check and execute would have needed the source's git grep (or a placeholder grammar in the stored text) at both, would have left the plan file itself stale between the edit and the next check, and would have had to leave alone the items a person edits into follow_ups: by hand. A plan written before this change keeps the line it was written with; plan again writes the new one.

The path itself is not in that line for the reason the line names keys: it would be a value the line cannot keep true. It is rendered where the mounts it follows from are rendered, at the moment they are printed. check computes it from the plan as it stands, execute from the mounts it made, so a hand edit of either key changes it, and it is never stored in follow_ups:. The reasons against re-rendering the stored line (the source's git grep at both readers, a stale plan file between the edit and the next check, the items a person writes into follow_ups:) do not apply to an additive sentence beside the mounts. It prints nothing where the spec leg holds no contracts/ (no contracts/ to lead to), and nothing where the two mounts are not two mount paths execute accepts (check has a finding for those, and a path computed from them leads nowhere).

Is contracts/path-classification.yaml a shape copy? No (L2-19)

• scripts/shape_materialize.py lists what a root carries: TEMPLATED, COPIED_VERBATIM, COPIED_FROM_SHAPE (repo_shape.py, validate-repository-naming.py, contracts/repository-naming.yaml) and the pin; assembly_root_paths() is those. The policy is in none of them, nor in FAMILY_*, update-shape.py, templates/ or the shape-pin template.
• adopt-project.py plan reads it from the shape's own tree (PATH_POLICY = SHAPE_ROOT / "contracts" / "path-classification.yaml"), shape-doctor.py from the standard it is run against.
• Measured: a real adoption's clone has 11 contracts/shape-pin.yaml rows (validate-pins.py, validate-manifest.py, bootstrap.py, Makefile, .gitignore, .gitattributes, validate.yml, AGENTS-shape.md, repo_shape.py, validate-repository-naming.py, repository-naming.yaml) and no tracked file naming path-classification.

So the change touches no root's pin: no root's next update-shape.py check sees a row for it, an upstream-changed row or an upstream-added one. Only the text of a plan written from now on differs. No template bytes and no --help text changed (scripts/render-cli-reference.py --check passes), so there is nothing to re-pin.

Tests

New: tests/test_adopt_leg_keys.py. Changed: tests/test_adopt_contracts_dir.py (the plan rows, spec_leg_from_code_leg's 11 rows back, and a leads-to test), tests/test_adopt_literal_names.py (the follow-up's one wording, follow_up(), and the two-argument call), tests/test_adopt_submodules.py (the harness follow-up now starts with the file's own path).

First round (L2-18, L2-19):

• test_a_hand_edited_mount_leaves_no_plan_time_path_in_a_follow_up (4 rows: the default layout and --spec-path legs/spec --code-path impl/code, each with --local-remote-dir and with a stub gh that logs the arguments of gh pr create): plan, edit both keys to elsewhere/s and elsewhere/c, run check and execute. check's follow-ups, the split commit's message and the body gh was given each carry exactly the one key-form line, and none of code/src/app/main.py, relative to `code/` or the plan-time path from the code leg to the spec leg. The legs are mounted at the edited paths. Since the second round it also asserts the sentence F1 adds, at the edited mounts (../s/contracts/).
• test_plan_names_the_read_across_by_the_two_keys (11 mounts): the line is the same text for every layout and holds neither mount path.
• test_the_contracts_rules_reason_names_no_leg_path and test_every_plan_entry_copies_a_reason_true_of_its_layout (default and nested plans): the reason holds no spec/, code/ or ../, and every spec-governance plan entry copies the policy's.

Second round (rv600's review):

• F1. With ACROSS_ROWS (default; nested legs/spec beside impl/code; both keys hand-edited, default to nested and back; the code key alone; the spec key alone), each path also checked to LEAD from the edited code leg to the edited spec leg's contracts/:
  • test_check_prints_the_path_to_the_contracts_beside_the_mounts (6 rows): check's mount line ends with it, computed from the plan as it stands;
  • test_the_split_message_prints_the_path_to_the_contracts_beside_the_mounts (6 rows, in-process): the same for _split_message's sentence;
  • test_execute_prints_the_path_of_the_mounts_it_made (3 rows, real execute): check and the split commit's message;
  • test_the_path_the_split_message_prints_leads_to_the_contracts: the printed ../../legs/spec/contracts/ followed from the code leg of a recursive clone of the hand-edited adoption reaches the spec leg's file;
  • test_the_spec_leg_is_named_relative_to_where_the_code_leg_is (11 mounts): adopt-project.py: a non-default --spec-path or --code-path is ignored by the adopted Makefile's CONTRACTS_DIR and by the cross-leg follow-up path #522's helper rows, restored;
  • test_the_path_is_printed_only_where_the_spec_leg_holds_contracts (5 rows: the directory, a file under it, nothing, other directories, a name that only looks like it), test_the_path_is_never_computed_from_values_that_are_not_two_mounts (6 rows) and test_check_prints_no_path_for_two_values_that_are_not_two_mounts (4 rows, real check): nothing is printed where there is no contracts/ to lead to or the mounts are not two mounts.
• F2. test_no_rules_reason_or_question_names_a_leg_path runs the LEG_PATH guard over every rule's reason: and question: and the default: block's (26 texts), and asserts it scanned every rule's reason; test_the_leg_path_guard_sees_a_mount_path (6 rows) holds the regex itself.

Counts (head d64252a on the merge a1a6973, /usr/bin/python3, git-filter-repo on PATH):

• The four whole files (test_adopt_leg_keys.py, test_adopt_contracts_dir.py, test_adopt_literal_names.py, test_adopt_submodules.py): 257 passed in 149 s, and 257 passed again under an uncommitted plugin that makes Path.write_text write CRLF (CI is Linux-only since ci: run the test suite on Linux only #597, so a CRLF bug would reach users unseen). The first round's "43 passed" and "31 failed, 12 passed" were not these four whole files: they counted the three files test_adopt_leg_keys.py, test_adopt_contracts_dir.py and test_adopt_literal_names.py plus the one changed test of test_adopt_submodules.py, test_a_submodule_takes_nothing_from_what_the_rules_imply (2 rows). The four whole files were 208 tests then (rv600 measured 208 passed on main + this PR, and 31 failed, 177 passed with the first head's tests over fbbbc8d's code).
• Targeted set, 49 files in two halves: tests/test_adopt_*.py (all), test_shape_doctor.py, test_scaffold_leg_path_collisions.py, test_cli_reference.py, test_repo_hygiene.py, every file naming cross_leg_references, follow_ups_for, path-classification or ../spec/contracts (test_doctor_next_is_a_command.py, test_stdout_encoding.py among them), and test_pin_source_unreadable.py, which scaffold-project.py and adopt-project.py: a pin records digest_source: local-clone and never the --pin-source path, so no committed pin file holds a local directory, and a --pin-source nothing reads is pin-source-unused (#540) #570 changed and which runs adopt-project.py: half A (23 files) 1883 passed, 3 skipped (20:32); half B (26 files) 2259 passed, 2 skipped (23:13); 0 failed, 4142 passed in all. The five skips: the three PowerShell rows of test_shape_doctor.py (powershell.exe exists only on a Windows host), test_adopt_plan.py::test_plan_on_medxehr_matches_the_rulings (no MedxEHR clone here) and a test_stdout_encoding.py row about Windows' own default encoding. scripts/render-cli-reference.py --check passes.
• Red on main 8f453d0, the four whole files of this head laid over main's code: 59 failed, 198 passed. The 59: 11 test_plan_names_the_read_across_by_the_two_keys rows, 4 hand-edit rows, 6 check rows and 3 execute rows (F1), the leads-to test, 6 _split_message rows, 5 holds-contracts rows and 6 never-computed rows (contracts_from_code_leg does not exist on main), the 3 reason rows and the every-rule guard (the policy's reason has ../spec/contracts on main), 11 test_adopt_literal_names.py rows (the new two-argument call is a TypeError on main, and the plan-level rows read code/src/reader.py names ...) and the 2 harness rows in test_adopt_submodules.py. The 198 pass on main: they pin behaviour main already has (adopt-project.py: a non-default --spec-path or --code-path is ignored by the adopted Makefile's CONTRACTS_DIR and by the cross-leg follow-up path #522's Makefile block, the symlink-staging rows, the end-to-end adoption rows) or nothing it can print (check's 4 non-mount rows).
• Red on the first round's head, the four whole files of this head over the merge a1a6973 (F1's code absent): 42 failed, 215 passed. The 42 are the F1 rows: 11 helper rows, the leads-to test, 4 hand-edit, 6 check, 3 execute, 6 never-computed, 5 holds-contracts and 6 split-message rows. F2 has no red row of its own: the policy is path-free, so its guard passes, and rv600's R4 below is how it is shown to bite.

Mutants of the second round, each applied once to a clone of d64252a (two clones; the four files that carry the tests, except R4's set) and reverted; every one is killed:

mutant	killed by
F1-1 check's mount line drops the path	13 failed: 4 hand-edit, 6 check, 3 execute
F1-2 the split message's sentence drops the path	16 failed: 4 hand-edit, 6 split-message, 2 holds-contracts, 3 execute, leads-to
F1-3 the helper climbs once whatever the code leg's depth (#522's auA-13 bug)	22 failed: 9 helper rows, 4 hand-edit, 3 check, 3 split-message, 2 execute, leads-to
F1-4 the path is computed from the spec leg to the code leg	20 failed: 4 hand-edit, 6 check, 6 split-message, 3 execute, leads-to
F1-5 no leg_path_problems guard	10 failed: the 4 check non-mount rows and the 6 never-computed rows
F1-6 printed whether or not the spec leg holds contracts/	3 failed: holds-contracts (nothing, other directories, a look-alike name)
F1-7 a file under contracts/ no longer counts	1 failed: holds-contracts a-file-under-it, and no other test
F1-8 check computes the path from the default mounts, not the plan's	14 failed: 4 hand-edit, 4 check, 4 check non-mount, 2 execute
F1-9 check swaps the two mounts	13 failed: 4 hand-edit, 6 check, 3 execute
F1-10 the split message asks the code leg's paths whether contracts/ moved	16 failed: 4 hand-edit, 6 split-message, 2 holds-contracts, 3 execute, leads-to
F1-11 the path loses its trailing slash	20 failed: 4 hand-edit, 6 check, 6 split-message, 3 execute, leads-to
F1-12 the sentence's own words change	20 failed: the same set as F1-11
F1-13 the split message takes the code mount for the spec mount	16 failed: as F1-10
F2-1 a rule's question: gains code/	1 failed: the every-rule guard
F2-2 the default: block's reason gains a mount path	1 failed: the every-rule guard
F2-3 spec-governance's reason keeps half the old text	4 failed: its own guard, the every-rule guard, both plan-entry rows
R4 (rv600's): code-source-and-tests' reason gains (../spec/contractsfromcode/)	1 failed, 553 passed, 4 skipped over rv600's set (the four files, test_shape_doctor.py, test_adopt_plan.py): the every-rule guard. It SURVIVED at 08b6b61 (502 passed, 4 skipped).

Mutants of the first round, each applied to a clone of the tree (the head's tests, before a one-line helper tidy in test_adopt_contracts_dir.py) and run against the four files that carry the tests; every one was killed. They were not run again at this head: no test that killed them changed:

mutant	killed by
M1 the old value form, with the default paths	28 failed: test_plan_names_the_read_across_by_the_two_keys, the four hand-edit rows, the literal-names rows, the harness row
M2 the line drops the legs.spec_path key	26 failed: the same, bar the harness row
M3 the line drops the file's name	28 failed: the same set as M1
M4 the file is named code/<path>	28 failed: the same set as M1
M5 follow_ups_for writes no read-across	24 failed: the plan rows, the hand-edit rows, the literal-names rows, the harness row
M6 the split message drops the follow-ups	5 failed: the hand-edit rows and #522's Makefile hand-edit test
M7 the PR body re-spells legs.code_path as the plan-time value	2 failed: the two stub-gh hand-edit rows, and no other test
M8 the reason restored	3 failed: test_the_contracts_rules_reason_names_no_leg_path, both test_every_plan_entry_copies_a_reason_true_of_its_layout rows
M9 the reason keeps ../spec/contracts from code/	3 failed: the same three
M10 the reason loses its point (reads across.)	1 failed: test_the_contracts_rules_reason_names_no_leg_path

README follow-up (after #519)

None. README.md:777 and the same row of docs/handbook.html describe the MedxEHR worked example, which is adopted with the default layout, where ../spec/contracts from code/ is true.

Closes #600

🤖 Generated with Claude Code

Summary by Sourcery

Keep adoption guidance consistent with the mounts actually used after a plan's leg paths are edited.

Bug Fixes:

• Keep read-across follow-ups accurate after users edit leg mount paths by naming the mount keys instead of stale plan-time paths.
• Include the current relative path to the spec leg's contracts/ directory alongside mount information in checks and split commit or pull request text.
• Remove layout-specific mount paths from the contracts policy reason so generated plan entries remain valid for nested layouts.

Enhancements:

• Add comprehensive coverage for hand-edited mounts, current-path rendering, invalid mount combinations, and path-free policy text across plan, check, execute, commits, and pull requests.

Tests:

• Expand adoption tests to validate follow-up wording, mount-relative contract paths, policy guards, and end-to-end behavior with local remotes and stubbed GitHub CLI interactions.

[sourcery-ai]

Sorry @brettheap, you've used your own review budget of 250,000 diff characters for the last 7 days.

You can request another review in 23 hours and 22 minutes by commenting @sourcery-ai review. Upgrade to get a review now.

[sourcery-ai]

Reviewer's Guide

The PR prevents stale plan-time mount paths from appearing in read-across follow-ups after leg mounts are hand-edited, updates the contracts policy rationale to be layout-independent, and validates the behavior across planning, checking, execution, commits, and pull requests.

Sequence diagram for stable read-across follow-ups after mount edits

sequenceDiagram
    participant Planner as adopt-project.py plan
    participant Plan as Plan file
    participant Operator
    participant Checker as check
    participant Executor as execute
    participant Commit as Split commit
    participant PR as Pull request

    Planner->>Planner: follow_ups_for(source, entries, names, collisions)
    Planner->>Plan: Store follow-up naming legs.code_path and legs.spec_path
    Operator->>Plan: Edit leg mount paths
    Checker->>Plan: Read _plan_follow_ups
    Plan-->>Checker: Layout-independent follow-up
    Executor->>Plan: Read _plan_follow_ups
    Plan-->>Commit: Same follow-up text
    Plan-->>PR: Same follow-up text

Loading

Flow diagram for layout-independent cross-leg follow-up generation

flowchart LR
    Source[Source paths and references] --> Cross[cross_leg_references]
    Entries[Plan entries] --> Cross
    Cross --> Follow[follow_ups_for]
    Follow --> Stored[Stored follow-up text]
    Stored --> Surfaces[Plan file, check, split commit, PR body]
    Mounts[Edited legs.code_path and legs.spec_path] -. do not rewrite .-> Stored

Loading

File-Level Changes

Change	Details	Files
Make read-across follow-ups independent of mutable leg mount paths by naming the plan keys instead of rendering plan-time paths.	Remove mount-path parameters from cross-leg follow-up generation and its callers. Change the follow-up wording to identify the source file in the code leg and reference legs.code_path and legs.spec_path. Remove the unused relative-mount-path helper and update affected tests and harness expectations.	adopt-project.py tests/test_adopt_contracts_dir.py tests/test_adopt_literal_names.py tests/test_adopt_submodules.py
Make the contracts governance rationale valid for arbitrary plan layouts.	Replace the default-layout-specific ../spec/contracts explanation with a layout-independent assembly-root statement.	contracts/path-classification.yaml
Add end-to-end coverage proving edited mount paths remain consistent across all follow-up surfaces.	Test default and nested plans after hand-editing both leg keys to unrelated mounts. Verify identical follow-up text in the plan, check, split commit message, and stubbed pull-request body. Verify the policy reason contains no layout-specific path and is copied consistently into plan entries.	tests/test_adopt_leg_keys.py

Assessment against linked issues

Issue	Objective	Addressed	Explanation
#600	Make the read-across follow-up independent of plan-time leg mount paths, so hand-editing legs.spec_path or legs.code_path does not leave stale paths in the plan’s follow-up, check output, split commit message, or pull request body.	✅
#600	Remove layout-specific mount paths from the contracts/ rule’s reason so it remains accurate for nested plans.	✅

Possibly linked issues

• Tracking: adopt-project.py, converting an existing repository in place #457: The PR implements both requested fixes: follow-ups name legs.spec_path and legs.code_path, and the contracts reason uses path-free wording.

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[brettheap]

(lane openRepoShape-2, 2026-10-15T23:10:27Z)

This lane reviewed its own PR #628 at 08b6b61, read-only in its own detached worktrees, and authored nothing.

MERGEABLE WITH NITS. There is no sequencing constraint.

Findings:

• P3, follow-up on Tracking: adopt-project.py, converting an existing repository in place #457: no surface now prints the relative path from the code leg to the spec leg (../spec/contracts/ for the defaults).
  • The key-form line at adopt-project.py:2013-2018 stays true after any edit of the two keys. check (:6196-6202) and the split message (:9183) print the current mounts beside it.
  • Unlike adopt-project.py: the Makefile block execute appends and the cross-leg follow-up name each leg where the plan mounts it, so --spec-path legs/spec exports a CONTRACTS_DIR that exists and a nested --code-path gets the right ../ path (#522) #572's Makefile line and adopt-project.py: the tracking branch defaults to the source's default branch, topics are set before the pull request, and NEXT names scripts/bootstrap.py when the shape's Makefile went aside (#479, #480, #481) #493-F3's tracking_branch line, its value appears nowhere in the same artifact.
  • A follow-up can add the relative path to those two sentences, computed from the plan's current keys when they are printed.
• nit: tests/test_adopt_leg_keys.py:223-232 checks only the spec-governance reason. A mutant that adds ../spec/contracts to the code-source-and-tests reason (contracts/path-classification.yaml:339-344) survives 502 tests. Check every rule.
• nit: in pr-body.md, "four files, 43 passed" and "31 failed, 12 passed" cover three files plus one test of test_adopt_submodules.py. The four files hold 208 tests.
• nit, follow-up: a plan written before this change keeps its old line, and check accepts it. That is right, because the line is free text. A sentence in the v1.1.0 notes covers an old plan whose keys were edited.

contracts/path-classification.yaml is not a shape copy, so no root needs a re-pin. We probed two hand edits of the keys, default to nested and nested to default, each with --local-remote-dir and with a stub gh. No plan-time path appeared in the plan, check, the split commit or the PR body. Seven of eight mutants were killed.

Tests:

• 48-file targeted set at 08b6b61: 3983 passed, 5 skipped.
• The four changed files on main plus adopt-project.py: the read-across follow-up names legs.code_path and legs.spec_path and no path, and the contracts/ rule's reason names none, so the hand edit of a leg's mount the tool asks for leaves no plan-time path in check, the split commit or the pull request (#600) #628: 208 passed.
• Under a CRLF write_text plugin: 43 passed.
• CI tests and SonarCloud are green on 08b6b61.

Merges: clean against main 8f453d0 and against #583, #594, #614, #617, #618 and #619, both pairwise and after #628.

The full report is at handoffs/xFactory/attachments/lane-openRepoShape-2/review628-08b6b61/final-report.md in opensoft/brett-wip.

[brettheap]

(lane openRepoShape-2, 2026-10-16T00:34:56Z) Head d64252a (rv600's F1 to F3, over a clean merge of main 8f453d0) was re-verified CLEAN (rv628r).

• The merge. a1a6973 is git's own clean merge, and this PR's diff is unchanged by it.
• F1. The plan still stores only the key-form line. check's mount sentence and the split message's mount sentence (and so the PR body) add the ready-to-type path, computed from the mounts they print.
  • 14 real plan, check and execute runs covered seven layouts, both remote kinds and hand edits of either key. Each gave the right path: ../spec/contracts/ by default, ../../legs/spec/contracts/ for legs/spec beside impl/code. Followed from the code leg of a recursive clone, each path reached the real file.
  • A brute force over 24,025 mount pairs gave no wrong path. The guard (leg_path_problems) prints nothing for any of the 1,309 pairs it refuses.
• F2. The guard now scans all 26 reason and question texts and asserts it saw every rule. rv600's mutant R4 now fails.
• F3. The counts in the PR body are now true. At this head the four files hold 257 tests; 59 are red over main 8f453d0.
• Tests: the four files give 257 passed. A merge with main f6376d7, run with the adopt tests adopt-project.py: check finds, and execute refuses before any leg, the branch names execute failed on, so adopt_branch: a..b, a source branch Adopt or one hidden by hideRefs no longer pass check and fail after both legs exist (#520, #420) #583 and adopt-project.py: execute hands git rm each path the split moves or drops as the bytes the tree holds, so a name outside ASCII under LC_ALL=C PYTHONUTF8=0 no longer stops the run with UnicodeEncodeError after both legs are pushed (#527) #614 changed, gives 932 passed.
• Not blocking. F4, an old plan's stale line, goes to lane 3's v1.1.0 release notes when this lands.

Waiting for a merge word from Brett Heap. d64252a needs Linux green (SonarCloud has passed).

[sonarqubecloud]

Quality Gate passed

Issues
4 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud