docs(audits): split hotcrm's 69,419 hand-written test lines three ways, and measure what os verify should have derived

[hotlong]

Part of #15418

An audit, not a code change: docs/audits/2026-09-hotcrm-handwritten-test-split.md.
No hotcrm file was touched, no test deleted or rewritten, no derivation built.

Part of rather than Fixes deliberately — the card's deliverable is the
measurement plus a record comment on the card, and the roadmap this produces
still has to be costed and dispatched. The maintainer closes it, not this merge.

The numbers

All 169 files under hotcrm test/ at 71a3452 were read and classified. Nothing
was classified by filename.

bucket	lines	share	~tokens
(1) platform should derive it, hand-written today	20,505	29.5%	129,752
(2) genuinely the app's own business judgement	28,242	40.7%	208,252
(3) grey zone — platform behaviour, app carries the cost	15,265	22.0%	97,972
(4) undecided	5,407	7.8%	30,610
total	69,419	100%	466,587

The subtraction the card asks for comes out at zero. os verify derives two
app-agnostic families — CRUD round-trip type fidelity (derive.ts + verify.ts)
and the RLS cross-owner invariant (rls.ts) — and neither corresponds to any
hotcrm bucket-(1) test. hotcrm also does not depend on @objectstack/verify at
all: it is absent from package.json, and deriveCrudCases / runCrudVerification
/ runRlsProofs return zero hits across the repository. The remainder is the
whole of bucket (1).

Three findings worth the reviewer's attention

1. hooks-runtime-service.test.ts (1,503 lines) is (2), not (3). The card's
   reason — "asserts what a hook does inside the platform's own sandbox" —
   describes action-sandbox.test.ts. This file imports the hand-written
   helpers/hook-harness, calls hook.handler(ctx) with the closure intact,
   and boots no sandbox and no engine. It is the app's own policy.
2. sharing-coverage.test.ts (1,309 lines) is 90% a docs test. Five of its
   eight describes check that the admin docs' OWD, sharing-rules and related-list
   tables match the app in every locale. The name reads as authorization.
3. Bucket (1) is 4.5x the card's at-a-glance guess (20,505 vs 4,558), and its
   two heaviest families — docs-to-metadata consistency at 9,851 lines and
   predicate totality at 3,415 — appear nowhere in the starting set.

Measurement basis

Token figures use hotcrm's own ratchet rule
(scripts/check-source-token-ratchet.mjs stripComments, blank-stripped,
chars / 4). Running that script unmodified reproduces its published
authored total ~135,830 for src, which is the check that the same rule is
being applied. test/ is 466,587 tokens by that rule — 3.4x the authored src
surface the "under 150k tokens" claim is measured on.

The honest reading requested by item 4 of the card: src plus the tests a
metadata app cannot avoid is ~344,000 tokens (135,830 + bucket 2), and that
floor already assumes the platform absorbs every line of bucket (3). It is 2.3x
the claim and does not fit one context window.

Of the 69,419 raw lines, 22,953 are comment and 5,947 blank — 40,519 execute.

Verification

node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack derived 16
runnable families for this diff; all 16 run green on the final commit 74c774f06,
through scripts/pm/os-verify-lock.sh.

check:doc-formula-expressions first exited 3 — its own "PREREQUISITE NOT MET,
nothing was measured" because @objectstack/formula and @objectstack/lint
were unbuilt in a fresh worktree. Both were built and it then reported its own
verdict lines rather than a skip: "self-test: 58 cases passed", "22
record-scoped formula example(s) across 427 files / 1371 TS blocks judged clean",
"9 @example(s) judged clean across 1159 packages/spec/src files", "14
predicate(s) ... judged clean; 6 skipped as undeterminable".

No changeset: this publishes nothing from any released package. docs/audits/-only
PRs carry none by precedent (76b6dd07e, 37ba31a4b, 1b51c683d — all
audit-only, all zero changeset files), so the skip-changeset label is applied.

🤖 Generated with Claude Code

https://claude.ai/code/session_01UHvF5hyiZjnCyExFnfQB8m

---

Generated by Claude Code