v0.57.0.

`cargo semver-checks` reports breaking library API changes: the squash
refactor (#2983) renamed `Repository::commit_subjects` to
`commit_message_details`, and the new `GitError::UnbornDefaultBranch`
variant (#2990) shifts the discriminants of the exhaustive `GitError`
enum. Neither is intentional public API, but together they force a minor
bump pre-1.0.

## CHANGELOG

### Improved

- **`wt step diff --branch`**: `wt step diff` gained a `-b`/`--branch`
flag, mirroring `wt step commit`, so the diff can target another
worktree's branch without leaving the current one. The branch must have
a checked-out worktree.
([#2995](#2995))

- **Squash templates can use commit bodies**: The squash commit-message
template gains an experimental `{{ commit_details }}` variable — a list
of `{ subject, body }` objects for the commits being squashed —
alongside the existing `{{ commits }}` (now documented as the commit
subjects). Templates can incorporate full commit bodies, not just
subject lines.
([#2983](#2983), thanks
@florianilch)

- **Recommended Claude Code commit command drops the `CLAUDECODE=`
prefix**: Claude Code removed the nested-session check that rejected
`claude -p` launched from inside another session, so the workaround is
gone. The recommended `[commit.generation]` command shown by `wt config
create` no longer carries a leading `CLAUDECODE=`, and `wt` no longer
strips `CLAUDECODE` from the environment before running
commit-generation commands.
([#2979](#2979))

### Fixed

- **Nushell wrapper installs where Nushell actually autoloads it**: `wt
config shell install nu` wrote `wt.nu` to
`$nu.default-config-dir/vendor/autoload`, which Nushell never autoloads
— on Linux the wrapper was written but silently never loaded, so `wt`
was never wrapped (it happened to work on macOS/Windows by coincidence
of path layout). It now installs to `$nu.vendor-autoload-dirs | last`,
and install/uninstall clean up any worktrunk wrapper stranded at the old
location. ([#2992](#2992),
thanks @nnutter for reporting)

- **Claude Code hooks work for Fish shell users**: The plugin's hook
commands used `${CLAUDE_PLUGIN_ROOT}` brace syntax, which Fish doesn't
expand; they now use `$CLAUDE_PLUGIN_ROOT`, so the activity and
worktree-lifecycle hooks fire correctly under Fish.
([#2962](#2962), thanks @amw)

- **Pager no longer wedges the terminal on Ctrl-C (Windows)**:
Interrupting the `--help` pager (`less`) with Ctrl-C on Windows could
leave the terminal in a broken state; `less` now quits cleanly on
interrupt. ([#2969](#2969),
thanks @ofek for reporting)

- **Clearer error when the default branch has no commits**: In a freshly
initialized repo whose default branch is unborn,
`merge`/`rebase`/`squash`/`push` (and the diff report) failed with
`Default branch main does not exist locally` plus a misleading hint to
reset the cached value. They now report that the branch has no commits
yet, without the wrong cache-reset suggestion.
([#2990](#2990))

- **`diagnostic.md` uploads as a gist again**: The `-vv` diagnostic
report inlined raw NUL bytes from NUL-separated git output, so `gh gist
create` rejected it as a binary file. Control bytes in the subprocess
preview are now escaped.
([#2991](#2991))

- **`wt list` tolerates a missing index file**: A repo with no
`<gitdir>/index` (nothing ever staged) made the temp-index probe fail; a
missing index is now treated as an empty one, matching git's own
behavior. ([#2884](#2884))

- **Inline code renders in `--help` section headings**: Terminal
`--help` showed literal backticks in headings authored with inline code
(e.g. the `wt config state logs` heading `Command log
(`commands.jsonl`)`). Headings now reduce inline code to plain text
under the heading's uniform style. The `--stage`/`--dry-run` subsection
headings in `wt step commit`/`squash` were also renamed to sentence case
("Staging", "Dry run").
([#3003](#3003))

### Documentation

- **`wt switch` docs give forge PR/MR URLs equal billing with
`pr:`/`mr:`**: The switch docs now present the full forge-URL form
alongside the `pr:N` shortcut.
([#2970](#2970))

- **New FAQ entry on moving uncommitted changes to a new worktree**.
([#3002](#3002))

### Internal

- **Bare-repo prompt opt-out stored as a hint**:
`worktrunk.skip-bare-repo-prompt` moved under the `worktrunk.hints.`
namespace, so it now lists under `wt config state hints` and clears with
`wt config state clear` (previously a top-level key that escaped both).
Clean cutover: users who already opted out are re-prompted once on their
next `wt switch --create` in a dotted-name bare repo.
([#3001](#3001))

> _This was written by Claude Code on behalf of max_

…em (#3602)

`wt config shell install` decided whether the file at a legacy location
was worktrunk's by reading it — a substring test for the wrapper header,
falling back to "every code line looks like an integration line".
Anything that didn't match was left in place. Fish sources `conf.d` at
startup, so whatever `conf.d/{cmd}.fish` defines is already loaded by
the time fish would autoload the `functions/{cmd}.fish` the install just
wrote — the stale definition wins and the new wrapper may never load.

Ownership is now the path. `conf.d/{cmd}.fish` and the stranded nushell
`{cmd}.nu` candidates are paths worktrunk itself computes for the
command name being installed, so install takes them back whole without
reading them. Only that exact filename is touched: a neighbour under
another name is not worktrunk's, and each removal is still reported.

`wt config shell uninstall` still reads the header, because it takes no
`--cmd` and so genuinely doesn't know the name — it lists the
shell-owned directories and has to tell our `{cmd}.fish` from the user's
own files beside it. That is the one place ownership can't come from the
path, and it prompts and previews every file before removing it, which
install does neither of. `is_worktrunk_managed_content`'s docstring now
says so; `is_worktrunk_managed_nushell` existed only to gate the
install-time deletion and is gone.

Audited the rest of the surface for the same principle: the OpenCode
plugin (`~/.config/opencode/plugins/worktrunk.ts`) is already removed by
path with no inspection, the Claude plugin delegates to `claude plugin
uninstall`, Gemini installs through `gemini extensions install` and
leaves nothing of ours on disk, and fish is the only shell with a
separate completion file — which uninstall already covers. No other
gaps.

## Data safety

This deliberately widens what install deletes, so it's worth naming: two
tests asserted the old conservatism and now assert the new boundary —
`test_configure_shell_fish_reclaims_conf_d_path` (was
`..._preserves_user_conf_d_file`, from #3589) and
`test_nushell_install_reclaims_only_the_command_name` (was
`..._keeps_unmanaged_legacy_file`, from #2992). Both put a neighbour
file in the same directory to pin that only `{cmd}.{ext}` is taken. The
FAQ's "What can Worktrunk delete?" inventory is updated to describe
path-based ownership rather than the old content-marker rule.

> _This was written by Claude Code on behalf of max_

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>