[NopeCharon]

What and why

collect_git_delta and collect_diff ran plain git diff, so the user's own diff configuration decided the format of the patch that parse_unified_diff renders in the Work attempt diff canvas, and that the Work page shows as raw text (details in #279):

• diff.mnemonicPrefix=true turns git diff HEAD prefixes into c/ / w/. The canvas names the file w/app.py and lists the real path again as a hunk-less "modified" entry.
• color.diff=always wraps every line in ANSI escapes. No header or hunk is recognised, so every file shows 0 additions and 0 deletions.
• diff.noprefix=true drops the a/ / b/ prefixes.
• diff.external replaced collect_diff's patch. Unlike the attempt delta, collect_diff did not pass --no-ext-diff.

This PR adds PATCH_FORMAT_ARGS (--no-ext-diff --no-color --src-prefix=a/ --dst-prefix=b/) to server/local_git.py and uses it for the committed and working-tree attempt patches and for collect_diff's patch. The options override the configuration for one command only, so the user's Git setup is left untouched. --name-only, -z and porcelain status calls are not affected by these settings and are unchanged.

Fixes #279

Linked Issue for product-semantic or public-contract changes: none (routine fix)

Change class

• Routine fix, documentation, test, maintenance, or presentation-only UI
• Product-semantic or public-contract change discussed in the linked Issue
• Isolated, default-off experiment

Owning layer: server/local_git.py, the Host helper that runs Git for machine consumers, and its callers in server/work_artifact_registry.py, which produce the attempt patch.

User-visible effect, or none: users with those Git settings see the correct files, hunks and counts in the attempt diff canvas, and a readable raw patch.

Compatibility or migration impact, or none: none. Patches persisted by earlier attempts are not rewritten. New attempts store the standard a/ / b/ format that the parser already expects.

Evidence

Commands and manual journeys run:

• python -m pytest -q tests/test_local_git_provider_diff.py tests/test_work_artifact_registry.py tests/test_unified_diff_canvas.py tests/test_work_export_integration.py: 48 passed (Linux, uv sync --locked --extra dev).

• The new tests test_attempt_patch_ignores_user_diff_format_config (parametrized for diff.mnemonicPrefix, diff.noprefix and color.diff=always, covering committed and working-tree patches through parse_unified_diff) and test_provider_diff_patch_ignores_user_diff_format_config fail on main and pass with this change.

• python -m ruff check server and the changed test files: passed.

• Full pytest tests on Linux (VN overlay tests needing tkinter excluded): 6115 passed, 84 failed. Re-running the failing files on unmodified main gives the same failures (ACP/Codex SDK, AUIP entry boot, Windows-only wallpaper and desktop tests), except two test_codex_approval_cancel.py cases that were timing-flaky under load; that file passes 14/14 on both branches when run alone.

• Relevant Python tests pass

• CPU/model-less baseline remains supported

• Electron npm run build passes when Electron code changed

• Dependency audit passes when dependencies changed

• Before/after screenshots are attached for visible UI changes

• Documentation/examples are updated for changed settings or contracts

Final check

• This PR addresses one coherent problem without unrelated cleanup
• It does not add a speculative API, fallback, or compatibility path
• No secrets, local state, model weights, voice material, or restricted assets are included
• Third-party notices and provenance are preserved

[Lucas1479]

Thank you for the focused fix and regression coverage. The shared patch-format arguments repair the four configurations reported in #279 at the Host Git-output boundary without changing user configuration. I independently checked real Git output for prefix/color handling and the external-driver boundary. All 15 hosted checks pass on this head. Additional submodule/copy/blank-context configurations are separate pre-existing follow-ups and do not block this scoped fix.