Doc references to the #410 scripts say NOT FUNCTIONAL, which is now stale

[realmarcin]

Found reviewing PR #522, which removes the five scripts that could never start (#410).

docs/AUTOMATION_TOOLS.md:204 carries a bare, unannotated instruction:

uv run python scripts/curate_evidence_with_pdfs.py --file FILENAME.yaml

No warning, no "not functional" marker — it reads as a working tool. It was already wrong before #522 (the script has never been able to start in any commit), and after #522 the file is gone as well.

docs/CURATION_PROGRESS_REPORT.md mentions the same script three times but each is prefixed # NOT FUNCTIONAL (#410):, so a reader is warned. AUTOMATION_TOOLS.md is the one that isn't.

Why this keeps happening: #410's guard, and the stronger one #522 replaces it with, both check scripts. test_no_working_script_points_at_a_script_that_cannot_run walks scripts/*.py looking at print/subprocess calls. Nothing walks docs/. So a dead pointer in prose is exactly the case neither test can see — and prose is where a curator is most likely to be reading, since they are following instructions rather than reading source.

This is the same shape as #442 (published pages still showing a corrected id): the fix landed in the code and the docs kept saying the old thing.

Fix applied in #522: AUTOMATION_TOOLS.md updated to name the working replacement (scripts/cache_fulltext.py then just validate-references), the CURATION_PROGRESS_REPORT.md markers changed from "NOT FUNCTIONAL" to "REMOVED", the stale .gitignore comment corrected, and a test added that walks docs/*.md for scripts/<name>.py references and requires the file to exist unless the line is explicitly marked as historical.

[realmarcin]

Correcting my own premise. The opening claim — that docs/AUTOMATION_TOOLS.md:204 is a bare unannotated instruction — is wrong. The warning is on the two lines directly above it:

# ⚠️ NOT FUNCTIONAL — this script imports communitymech.literature_enhanced,
# which has never existed, so it fails at import. Tracked in #410.
# (The command below is also stale: this repo uses uv, not poetry.)
uv run python scripts/curate_evidence_with_pdfs.py --file FILENAME.yaml

I grepped for the script name, got the line number of the command, and read that line alone. Every doc reference is in fact annotated. This is the third time this session a line-scoped text scan has missed the prose that negates the hit — the same trap as grepping a curation note for the id it retired.

What is actually left, and it is much smaller: after #522 the scripts are removed, so "NOT FUNCTIONAL … Tracked in #410" is stale in a new way. A reader now looks for a file that isn't there and cannot tell whether it was deleted or they are on the wrong branch. The markers should say REMOVED, name the replacement, and stop pointing at #410 as if it were open.

Still worth keeping from the original: no test walks docs/ for references to scripts/<name>.py. Both #410's guard and the stronger one in #522 check scripts/ only. That gap is real regardless of my misreading, and it is why the staleness here can only be caught by hand — so the test is still worth adding, just for "the file exists" rather than "the instruction is unwarned".