[ea-open-source]

Chore #47472 (/s/…) said the README named one deps divergence where CONTRACT.md named two. Both counts were stale. Both server asks the deps section cites shipped on 2026-08-06, and production runs them (probed against api.eastagiletracker.com on 2026-09-10).

#35639 shipped, and the ordering divergence is gone

CONTRACT.md predicted "a settable order on the public create, and a tiebreaker on the projection". CreateBlocker never gained an order field. It is still { blocker_desc, resolved }.

agile-tracker a774ed013 (PR #1521) made blockers.rs::create assign the order server-side instead:

INSERT INTO blocker (story_id, blocker_desc, resolved, blocker_display_order)
SELECT $1, $2, $3, COALESCE(MAX(blocker_display_order), -1) + 1 FROM blocker WHERE story_id = $1

Both story projections in stories.rs now ORDER BY story_id, blocker_display_order, blocker_id, so the tiebreaker shipped too. Three sequential public creates returned blocker_display_order 0, 1, 2 in production and listed back in that order.

The direct-engine writer already posts blockers one after another in GitHub's blocked_by order, so its rows land at the same positions the server importer writes from the entry index. Order parity holds.

• CONTRACT.md drops the blocker_display_order divergence bullet and states the parity beside the Selection rule, where the listing order already lives.
• The test stops being a divergence test. It keeps the "nothing order-shaped to send" assertion, adds the blocked_by order assertion, and pins the server half with a hand-port of the create's order assignment, the same contrast pattern the file already uses for description back-links.

#35629 shipped toward characters, and left the CLI over-strict

CONTRACT.md assumed the public route was the stricter end. agile-tracker 8e5ed5479 (PR #1509) changed validate_length to count characters, not UTF-8 bytes. The importer is unchanged at relations.rs:502, blocker.desc.chars().take(255).

Both server ends now agree at 255 characters. Production accepted a 255-character / 765-byte blocker_desc and rejected 256 characters with 400 too_long.

The CLI still cuts at 255 bytes. The divergence survives, and its cause inverted: CONTRACT.md's claim "The CLI's behaviour is the only one the public route accepts" was false.

• The length bullet now says the CLI clamps in bytes and both server ends clamp in characters, notes that #35629 reconciled the two server ends, and cites the live companion for the CLI half.
• The CLI fix is out of scope here. It is filed as project-91 bug #482645 (/s/3zktusgn). No src/ file is touched.
• CONTRACT.md:1520 and CONTRACT.md:2003 repeat the same stale str::len() rationale. They are inventoried on #482645 and stay untouched here.

The knock-on at the Text bullet ("the first divergence listed below") and the section header ("two write-side divergences") are corrected to one.

The README

(what the public API accepts) was false, so it is gone. The bullet keeps "one narrow exception", describes the effect in user terms rather than by column name, states the order parity in one clause, and cites the live #482645 instead of the closed #35629.

The gate gap the story names as the root cause

#35639 landed in CONTRACT.md and in tests/parity.test.js but never in the README, because the meta-test that guards divergence citations reads CONTRACT.md alone.

A new test reads both files:

• It collects the live companion ids CONTRACT.md cites in its write-side divergence bullets, in the #NNNN (/s/slug) form, and asserts README's deps bullet cites every one.
• It also rejects the rot the other way: an id in README's deps bullet that no divergence bullet names as a live companion. That is what catches a README still pointing at a closed ask.
• Both halves anchor on stable markers, not line numbers, and the failure message names which id is missing from which file.

Red proof. Against the un-edited docs at origin/main:

AssertionError [ERR_ASSERTION]: CONTRACT.md names "`blocker_display_order` is not settable
from the CLI." citing #35639, README.md's `deps` bullet cites none of them

Against the edited CONTRACT.md with the README reverted, and again with the README citing the closed #35629:

AssertionError [ERR_ASSERTION]: CONTRACT.md names "Length is clamped in bytes by the CLI,
in characters by both server ends." with companion #482645, README.md's `deps` bullet does
not cite it

Gates

Gate	Result
npm test	tests 1248 / pass 1247 / fail 0 / skipped 1 — rc=0
npm run lint	Found 1 warning. Found 1 info. (both pre-existing on main) — rc=0
npm run typecheck	clean — rc=0

Docs and tests only. src/ is untouched.

🤖 Generated with Claude Code

https://claude.ai/code/session_011THnzMbQzbJj7mw2xxHs8i