docs(readme): drop the docs/ file count, which was true once and cannot stay true

[wshallwshall]

What

docs/README.md line 3 opened with "There are 377 markdown files under docs/". The real
count is 285. The sentence now carries no number.

-There are **377 markdown files** under `docs/`. Most of them are maintainer planning history, not
-documentation. This page exists so you do not have to guess which is which.
+Most of `docs/` is maintainer planning history, not documentation. This page exists so you do not
+have to guess which is which.

Where 377 came from

It was correct the day it was written, and has never been updated. dc3a7630b added the index on
2026-07-31; its parent 304a93b85 carries exactly 377 markdown files under docs/. The author
counted the tree before adding the index, so the claim shipped off by the one file being written.

The count steps rather than drifts. Measured through the GitHub API at each of the six commits that
have touched this file:

commit	date	actual docs/**/*.md
304a93b85	2026-07-31	377 (parent; the number written)
dc3a7630b	2026-07-31	378
fdec72cac	2026-08-06	412
123d08b65	2026-08-06	311
921db74a1	2026-09-01	280
b9f4f3810	2026-09-04	285

The tree lost about a hundred markdown files between 2026-08-06 and 2026-09-01. Two of those six
commits are ADR 0160 untracking passes, which move maintainer material into the vault clone.

Two hypotheses were checked and both fail. The primary checkout holds zero untracked or
gitignored .md under docs/, so 377 did not count vaulted files a plain find cannot see. The
vault clone holds 585 .md under docs/, which is not a 92-file top-up either.

Why no number and no pinning test

A pinned count would assert a figure whose meaning depends on which checkout reads it, inside a
file whose very next paragraph says docs/ is deliberately partial. The sentence's job is to tell
a reader that most of docs/ is maintainer planning history and that the index exists so they need
not guess which is which. That survives without a count.

Deliberately not touched: docs/BACKLOG.md line 6912 records "347 markdown files" as part of a
dated re-measurement inside a shipped item. A dated measurement says when it was true, so it is a
record rather than a claim that rots. No ledger row is needed for this change and none is added, so
this does not contend with the nine open PRs queued against BACKLOG.md.

Checks

Run locally, all green:

• pre-commit run --files docs/README.md -- ledger gate, forbidden-content, control characters,
  username-as-access-key, gitleaks all Passed; the Python and workflow hooks skipped, no such files.
• python scripts/docs/link_check.py docs -- 4847 relative links across 285 markdown files, every
  one resolves.
• pytest tests/test_dangling_citation_check.py tests/test_link_resolution.py tests/test_ech_record_premise.py -- 75 passed.
• pytest tests/test_private_paths_stay_ignored.py -- 35 passed (it is the only test file that
  mentions docs/README.md, in comments about the "Start here" section, which is untouched).

Not run, and not applicable: ruff, mypy, and the rest of pytest. No Python changed.

One instrument note for a reviewer repeating this: link_check.py docs/README.md reports
0 relative links in 0 markdown files and exits 0. It takes a directory, so the file form is a
false green. The run above passes docs.

Full history is not reachable from a local clone here -- this worktree and the primary are both
depth-1 shallow, where git log -S and --diff-filter=A name the single commit for every string
and every file. The table above came from the GitHub API instead.

Co-Authored-By: Claude Opus 5 noreply@anthropic.com

🤖 Generated with Claude Code