docs(readme): bind the crate count to the workspace

[seonghobae]

Defect

The README had two inconsistent workspace crate-count claims while the measured workspace contains 58 directories under crates/ with a Cargo.toml. This Draft binds every explicit README count and the adjacent crate inventory to the repository instead of relying on manual refresh.

Current repair

• Both README claims are 58.
• tests/quality/test_readme_crate_count.py requires at least one explicit N independently documented ... crates claim, checks every such claim against the live workspace count, and requires the README crate list to equal the workspace crate set.
• docs/product-technical-gap-baseline.md is not touched; that shared ledger remains owned by Draft docs(ddd): restore queue authority and bounded-context ownership #435.

Current exact stack

Current exact head: 21e0e12691efc7eccd22ab832b83b978ee271e43.

This dependent Draft is now stacked on the current main-target foundation successor #538 exact 0a6947b11f2dba9e161f2abe8bdec0f652d4b47a (fix/contradictory-zero-count-records) rather than the stale #523 base.

The ordinary-forward restack commit 21e0e126... has prior #531 exact be17b82967c7dc0f752e35aafa008b33fc1f6a5f as first parent and #538 exact 0a6947b... as second parent. The branch ref advanced with force=false; no destructive rebase or force update was used. Relative to #538, the intended surviving delta is only README.md plus tests/quality/test_readme_crate_count.py.

Evidence boundary

Pre-restack local evidence and predecessor-head checks do not transfer to 21e0e126.... Keep this PR Draft while #538 is unlanded. After #538 survives protected-main integration, restack this unchanged delta non-force onto the surviving protected head if necessary and reacquire applicable exact-head checks plus qualifying independent review.

Do not self-approve, bypass required gates, copy the shared product-gap ledger into this lane, or treat skipped/cancelled/predecessor checks as GREEN.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

The count was right and the list beside it was still wrong

Correcting the two numbers was not enough, and I should have checked this in the first pass rather than trusting that a count and the list under it agree.

The README also prints the crate list. It held 56 of the 58 directories. mlx_native_receipt and orchestrator_live were absent, so the corrected count sat directly above a list that disagreed with it. A reader checking which crates exist would have missed both.

listed in README: 56   actual: 58
on disk but missing from README: ['mlx_native_receipt', 'orchestrator_live']

Both are now listed, in sorted position.

test_readme_lists_every_crate compares the printed list against crates/ as a set, so a stated count can no longer stand beside an incomplete list. Deleting one entry fails it and names the missing crate:

AssertionError: Items in the second set but not the first:
FAILED (failures=1)

Merged, not forced

Another session restacked this branch onto #523's successor while I was working. I fetched and merged rather than forcing, and both changes stand together with no conflict.

Gate	Result
python3 -m unittest discover -s tests/quality -p 'test_*.py'	185 tests, OK
python3 -m coverage report --fail-under=100	100% (1380 statements, 694 branches, 0 missing)
python3 scripts/validate_documentation.py	passed
grep -c '^crates/' README.md	58
git diff --check	clean

Note that this branch now inherits #523's draft guard through that restack, so its merge is coupled to the policy question open on #523. That coupling was not mine to make, and I have left it as the other writer arranged it rather than unpicking someone else's restack.