[VasiliyRad]

Why

The [file:|send:|attach:] marker regex is duplicated inline at three sites in discord-bridge.py:

Line (pre-PR)	Path	Used by
2832	poll_results	channel replies
2974	poll_proactive	owner DMs
3130	poll_dm_fallback	channel-redirect branch

The pattern requires absolute paths (/... or ~/...) — relative paths previously resolved against CWD, which differed between launchd-managed and bare-shell runs. Any future hardening to the regex would need to land in all three places or one path would drift.

Changes

• _FILE_MARKER_RE — module-level compiled regex.
• _split_file_markers(text) -> (clean_text, files) — pure helper. Three call sites now call it instead of re-defining the regex.

No behavior change. Refactor is byte-for-byte equivalent for valid inputs; the only difference is that the regex compiles once at module-load instead of per polling-loop iteration (minor perf win).

Tests

tests/discord-bridge-file-markers.test.py — 10 cases pinning the contract:

Case	Pins
no markers / empty / single	Common cases.
file / send / attach aliases	All three keywords.
~/... allowed	Home-relative paths.
relative paths rejected	Pre-fix regression guard.
multiple markers preserve order	Textual order for the send loop.
[reply: 12345] does NOT match	The reply directive uses a different keyword.
unknown keywords ([path:] etc.) don't match	Strictness.
Architectural — regex body appears in source exactly once	A future refactor that re-introduces an inline copy fails here.

All 10 pass locally.

Test plan

• python3 tests/discord-bridge-file-markers.test.py — 10/10 pass
• Syntax check
• Manual: send a reply with a [file: /tmp/sutando-x.png] marker → verify file delivered as attachment, marker removed from text

🤖 Generated with Claude Code

[CLAassistant]

All committers have signed the CLA.

[sonichi]

LGTM. Refactor with a structural test, which is the right shape for "consolidate inline duplicates" — without the architectural assertion that the regex body appears exactly once in the source, this PR would just be cosmetic and the drift recurs on the next change.

All three call sites converted (poll_results, poll_proactive, poll_dm_fallback). The relative-path rejection is explicitly tested — important because that's the actual constraint distinguishing this regex from a casual file-extraction pattern (launchd CWD ≠ bare-shell CWD makes relative paths silently break for proactive DMs).

[reply: 12345] correctly stays excluded (different keyword), and the test covers it — easy to break by mistake if someone "generalizes" the file-marker keyword set later.