Files
Evil0ctal fd61eff1ad chore: put the mark back, and take the salvage tree out of the repo
Three tidies, ahead of this branch ever being published.

**The mark.** The cat has been at the top of this project's entry points
since v1 and did not survive the rewrite. It is back on the five files
people actually open first: the package root, the API app, the CLI, the
worker and the MCP server.

It needed an exemption, because it is drawn partly in katakana and two
separate tests assert this codebase's Python contains no CJK. The
exemption is by SHAPE, not by filename: a line passes only if every
character in it comes from the mark's own small alphabet, so a line that
smuggles a sentence in beside the cat still fails. Both checks now share
one predicate in tests/support/marks.py rather than keeping a copy each -
a rule that exists twice is a rule that will eventually disagree with
itself.

**vendor_salvage/ is no longer tracked.** Nothing imports it; it was V4
source kept for reference while v5 was being written, and .gitignore has
listed it since before it was committed - which does nothing for a path
already in the index, so it has been shipping regardless. It stays on
disk locally. It also held a byte-identical copy of the GPL-3.0 A-Bogus
port, so this is the second half of the licence fix that landed with the
re-derivation.

The six docstrings that cited paths inside it now cite the upstream
commit instead (branch main, 8c98fb7). That is better provenance than it
was: a reader can fetch a commit, and could never fetch a directory that
only existed in one working copy.

**adm_settings.json is gone.** A captured API response envelope - 53
settings, success/data/error/meta and all - committed by accident in
6e7d862 and referenced by nothing. Checked before deleting: the two
sensitive-looking keys hold an empty list and an empty string, so nothing
leaked.
2026-09-09 21:20:56 -07:00
..