[dmccoystephenson]

Summary

The restricted-fetch-refspec entry in the generated-skill template's Edge cases section is extended to cover the inbound symptom reported in #130. The entry is retitled so that it can be found from either error message.

• Outbound (unchanged): a branch that has just been pushed fails gh pr create auto-detection, and the fix is --head <branch>.
• Inbound (new): git fetch origin <branch> writes only FETCH_HEAD, so git checkout <branch> fails with error: pathspec '<branch>' did not match any file(s) known to git for a branch that does exist on the remote. git fetch origin +<branch>:<branch> is prescribed from the default branch.
• Beyond the issue's suggestion: a second hazard was observed in this clone. Local branches from earlier runs are kept (ten were present at triage), so a plain git checkout <branch> can succeed silently onto a stale local copy. Two checks were added: git rev-parse <branch> is compared with the PR's headRefOid, and git log FETCH_HEAD..<branch> must be empty before the forced update, so that unpushed work is not discarded. The + was added to the issue's suggested <branch>:<branch> form for this reason.
• The entry is kept as one line, matching its siblings. A dash list would have caused the following {{#if REVIEWER}} entry to be read as a lazy continuation of the last list item.

Closes #130

gardener interface: Steps 3, 5, and 6 are untouched, so no breaking change is introduced for gardener.

Retrofit: the Edge cases section is copied into every generated skill, so a retrofit pass is likely needed for every existing <slug>-dev-loop skill that carries the old gh pr create-only entry. create-dev-loop-dev-loop is one of them.

Research grounding

No RESEARCH.md finding applies. This is plain Git refspec behavior, as stated in #130, so no Implementations entry is added.

Doc sync check

• README.md's "What it does" Step list still matches create-dev-loop.md's Steps 1:1 (no Step was changed)
• No {{placeholder}} was added. The existing {{DEFAULT_BRANCH}} is not used in the new text.

Test plan

• python3 scripts/check_docs.py: "Doc consistency check passed."
• python3 -m unittest discover -s tests: 42 tests were run, OK.
• The inbound symptom was reproduced in this dispatch's clone (refspec +refs/heads/main:refs/remotes/origin/main, git 2.25.1). git fetch origin fix/negated-closing-keywords wrote only FETCH_HEAD, and the checkout failed with the quoted pathspec error.
• The stale-branch case was observed: while a local fix/negated-closing-keywords was present, git checkout succeeded without any tracking message. git fetch origin +fix/negated-closing-keywords:fix/negated-closing-keywords was then run against a local branch reset to 855a4b7, and it reported (forced update) to ec84115, the remote tip. The scratch branch was deleted afterward.
• UNVERIFIED (behavioral): no /create-dev-loop run was performed, because a GitHub repo is created in Step 6 and files outside the checkout are written in Steps 3 and 5, and neither is authorized in a headless dispatch ([self-audit] The repo's own validation anchor is structurally unrunnable in a headless dispatch, so template changes can never merge autonomously #109). None of the five CLAUDE.md checklist items was exercised, so this PR is left for a maintainer to merge.

Issues not picked this cycle

• CLAUDE.md still links gardener under the old dmccoystephenson owner #106 and the CLAUDE.md half of Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173: explicit authorization is required for CLAUDE.md edits.
• [self-audit] The repo's own validation anchor is structurally unrunnable in a headless dispatch, so template changes can never merge autonomously #109: the options listed there are left as a maintainer decision.
• Unverified: whether an in-checkout .worktrees/ survives gardener's refresh (git clean -fdx) #114: an observation of gardener's refresh is needed, which cannot be made in this dispatch.
• Phase 1 repo resolution assumes the configured clone is clean; stale staged work in a shared cache clone blocks git pull #148: Phase 1's resolution block is also touched there, but a shell-variable and for-loop rework (per its comment) is needed, which is larger than this one-entry fix.
• All other open issues (Phase 4's rubric scores the diff only, so PR-description obligations like CLAUDE.md's retrofit note go unchecked #120, gh prose examples carry no attribution sign-off, and --body-file paths must be literal (a shell variable is invisible to the hook) #122, The >50-deleted-lines do-not-auto-merge heuristic misfires on extraction refactors, test migrations, and generated artifacts #124, grep is not on tend's allow-list, so generated validation checklists and verify-by-removal assertions cannot run headlessly #127, An approval-gated CI run reads as 'no checks reported', which is neither the green nor the red Phases 1 and 4 branch on #129, Test the mechanism, not the stochastic outcome, when a mock cannot exercise an acceptance criterion #131, Phase 4 verifies backward 'Closes #N' references but not forward 'a follow-up issue has been filed' promises #133, Tests-fix has no path for a scope-blind anchor, only for an anchor that cannot run #140, Upstream COMMANDS.md is trusted as source, but it can disagree with the command class it documents #141, Stage A sweeps fix the instances named, not the class — and a prose grep is treated as a CI-strength anchor #143, Phase 1 triage never checks whether the default branch's own CI is green #144, A refused environment probe is not grounds for UNVERIFIED — the anchor has to be attempted first #145, Phase 2: bundle a small docstring/annotation fix with a co-located test or doc update as one PR #146, Phase 3: never pre-escape Edit-tool strings — verify the diff for literal backslash-quotes after inserting quoted text #147, Tests-fix revert produces compile errors, not failures, when the new tests reference symbols the fix introduces #149, Phase 5: compress the review wait for a single autonomous run with no reviewer assigned or requested #150, Dispatched runs can load an installed skill copy older than the rules merged in the skill's repo #158): deferred, unassessed in depth, so that this PR is kept to one theme.

This PR description was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

🤖 Generated with Claude Code

---

drafted by Claude on behalf of Daniel Stephenson

[dmccoystephenson]

Self-review rubric (an inline review was performed in a headless dispatch, with no independent reviewer):

• Scope: PASS. One line of create-dev-loop.md (the restricted-refspec Edge case) was changed, and nothing else.
• Tests-new: PASS (not applicable). No function was added.
• Tests-fix: PASS (not applicable as a code regression test). The issue describes missing guidance, not a code defect. Its symptom was reproduced in this clone, and the prescribed +<branch>:<branch> form was run and observed to produce (forced update) to the remote tip.
• Sibling structure: PASS. The entry is kept as one bold-titled line, like its siblings. A dash list was tried and removed, because the next {{#if REVIEWER}} line would have been absorbed into it as a lazy continuation.
• Sibling renames: PASS. The entry title was changed, and a repo-wide search for must first push and pathspec found no other references.
• Docs: PASS. The README, CONTRIBUTING, SECURITY, and template docs do not describe Edge-case entries individually.
• Issue resolution: PASS. Both changes suggested in A restricted fetch refspec also blocks checking out an existing remote branch, with an error that reads as 'branch does not exist' #130 are present: the inbound symptom is added to the existing entry, and the entry is retitled so it can be found from either error message.
• CI: PASS. doc-consistency passed on the PR head.
• Placeholder parity / README Step parity / fence escaping / phase numbering / version comments: PASS. No placeholders, Steps, fences, phases, or header lines were touched (check_docs passed).
• Research grounding: PASS. The PR body states that no finding applies.
• No back-ported specifics: PASS. The text is generic to restricted clones. "gardener-managed" was already in the entry.
• gardener interface: PASS. Steps 3, 5, and 6 are untouched, and this is stated in the PR body.
• Retrofit note: PASS. The PR body names the existing skills that likely need a retrofit pass.
• Behavioral validation: FAIL (UNVERIFIED). No /create-dev-loop run could be performed headlessly ([self-audit] The repo's own validation anchor is structurally unrunnable in a headless dispatch, so template changes can never merge autonomously #109).

Judgment calls flagged for a reviewer:

• create-dev-loop.md:584 — the stale-local-branch clause goes beyond A restricted fetch refspec also blocks checking out an existing remote branch, with an error that reads as 'branch does not exist' #130's suggested text. It is grounded in the ten leftover local branches observed in this clone at triage. A reviewer may decide whether it belongs here or in a separate entry.
• create-dev-loop.md:584 — the + forced update can discard local-only commits. The git log FETCH_HEAD..<branch> precheck is meant to prevent that, but the sequence (plain fetch, then log, then forced fetch) was not run end-to-end against a branch that had unpushed commits.

Because of the UNVERIFIED behavioral item, this PR is not merged autonomously and is left for a maintainer.

This review was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson

[Copilot]

🟡 Changes recommended

The unconditional git log check fails when the local branch is absent, blocking the intended recovery; it must run only for existing branches and succeed with no commits before allowing a forced update.

1 open issue

• Skip local-commit check when the branch is absent · New

What changed in this PR

Extends the generated skill’s restricted-fetch-refspec guidance to cover inbound branch checkout failures, alongside the existing PR-creation workaround.

Changes:

• Retitles the Edge cases entry to make both error messages discoverable.
• Adds explicit fetch instructions and checks for stale local branches and unpushed commits.

File	Description
create-dev-loop.md	Expands restricted-refspec troubleshooting in the generated skill template.

_{Review effort: Balanced}

[dmccoystephenson]

Copilot's finding ("Skip local-commit check when the branch is absent") was confirmed and addressed in 98bfb83.

• The failure was reproduced: with no local branch, git log FETCH_HEAD..<branch> exits with fatal: ambiguous argument ... unknown revision.
• The unpushed-work check is now prescribed only when git rev-parse --verify --quiet <branch> prints a SHA. That probe was observed to print nothing for an absent branch and a SHA for main. When no local branch exists, the forced fetch is run directly.
• python3 scripts/check_docs.py passed, and the unit tests passed (OK).

The behavioral /create-dev-loop validation remains UNVERIFIED (#109), so the PR is still left for a maintainer to merge.

This comment was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson