docs(releases): draft the 17.6.0 release notes and upgrade checklist

[hotlong]

What this is

A pre-cut draft of the curated release page for 17.6.0, content/docs/releases/v17/17-6.mdx, plus its entry in content/docs/releases/v17/meta.json and in scripts/docs-audit/handwritten-docs.json. It also appends one dated correction to content/docs/releases/v17/17-5.mdx (details below).

Docs-only. Everything is under content/docs/releases/ (release-owned) or the docs-audit list, so this is the dedicated docs-only PR AGENTS.md sanctions for that tree. It publishes nothing from any package, hence skip-changeset.

Clause-②: no

It follows the 17.5.0 page's structure:

1. Highlights
2. What's new, including the "read this before treating the version number as a safety guarantee" list of silent runtime changes
3. Breaking changes & migration (triaged, not exhaustive)
4. New capabilities
5. Notable fixes
6. New in Console (Studio)
7. Shipped in 17.5.0, listed again in 17.6.0's CHANGELOG
8. Upgrade checklist (every line marked not exercised)

How it was compiled

• Source: all 338 changesets pending on main at 748b2407. Every changeset was read and triaged into breaking, feature, fix or internal. PR numbers and short SHAs are taken from the commit that added each changeset.
• Already shipped in 17.5.0: 16 of the 338 changesets (from 15 commits) describe code that 17.5.0 already published. I checked each commit with git merge-base --is-ancestor against the version commit 8c87d26a and the publish commit 0f6dcac5.
  • Eight follow the version commit in first-parent order. The 17.5.0 page already covers them.
  • Seven are ancestors of the version commit but were not in the version PR when it merged. The 17.5.0 page does not mention them. Two of them are breaking (e73ee2d feat(spec)!: RealtimeEventType names the emitted data.record.* / data.records.* vocabulary (#20288) #20567, c876a74 fix(spec,driver-turso)!: refuse a forced mode replica with no syncUrl at authoring and at construction (#20437) #20504), and 7a1faf1 (fix(spec,cli): os validate / os build read the ADR-0087 conversions defineStack applied — --json conversions and --strict see the producer's record #20579) makes os validate --strict fail on a live conversion.
  • The new section matches the list in .changeset/20622-release-aftercare-upgrade-and-unannounced-notes.md.
• Cross-check: each of the nine ADR-0087 conversions added since 17.5.0 is covered by a migration entry on the page:
  • action-block-endpoint-to-target
  • connector-sync-keys-removed
  • connector-triggers-removed
  • cube-refresh-key-removed
  • dataset-count-measure-empty-field-removed
  • form-field-public-picker-removed
  • form-view-subform-columns-canonicalized
  • page-header-breadcrumb-removed
  • time-default-utc-suffix-dropped
• Console section: built from the three pin-bump changesets (dd3f7e1be356 → db11afd4967c → e420df310f5b → 31971ff1e28f), which carry 232 releasing objectui changesets, 23 of them declared breaking. The range was also read with scripts/objectui-range.mjs.
• Fact-check pass: a second pass checked every cited SHA's changeset against the page, in three slices. It found 31 claims that were wrong or overstated, and the three follow-up commits correct them. Some examples:
  • An RLS contains verdict that d2bc644 (fix(core,objectql,plugin-security): the RLS write check judges a lone scalar on a declared multi-valued field as the list it is stored as #21253) later reversed in this same release.
  • The manage_platform_settings scope: reads of both types, but writes of datasource only.
  • The dataset door's status after 434c6c7 (feat(spec)!: an analytics dataset dimension's and measure's field is a column reference (#21220) #21240).
  • Filter positions that differ per change.
  • 7a1faf1, which is not a breaking change.
• Other corrections made while compiling:
  • The connector migration table does not claim that a job schedules a connectorSource pull. 0efbdc3 says nothing schedules a pull until the job stage lands, and none landed in this release.
  • The public-form picker search-key change (bafb8c9, fix(rest): the public lookup picker searches and sorts by the first display field the caller may query #21136) is noted as moot, because the route it changed is retired later in the same release (3dc33b2, feat(spec,rest,lint)!: retire the form field's publicPicker and the anonymous lookup route (#21180) #21222).

Open items for the cut

These are recorded in the page's RELEASE-TIME TODO comment.

• Anonymous endpoints. Under the deny baseline (fix(plugin-security,spec)!: a non-system caller that carries a principal and resolves no permission set gets the deny baseline at object admission and at the row scope #21217), an app-declared anonymous endpoint (authRequired: false) can no longer read or write objects. security(core, plugin-security): an unauthenticated principal never resolves the permission sets bound to the guest anchor; ADR-0090 D9 is declared and seeded but not enforced #21158 is still open, and until it lands no supported channel grants anonymous callers a permission set. The page says so instead of prescribing a grant.
• Console pin. Three 17.6.0 server changes refuse a body that the pinned Studio (31971ff1e28f) still sends. objectui fixed each one after the pin:

Studio action	Server change that refuses it	objectui fix (not in the pin)
A dataset count measure saved with the Field box blank (field: '')	#21240	0858267e
An External / Validate-only datasource saved without a credential	#21133	8001068b
Republishing an html page that gained a plugin component	#20852	3ae91930

The page lists these under "Known console issues". If the pin moves before the cut, the lines it fixes come out. Otherwise the accepted-for-GA waiver is recorded.

The correction to 17.5.0

One dated correction line is appended in place to the "Also shipped in 17.5.0" paragraph of 17-5.mdx, in the form #20985 used. It names the seven commits that shipped in 17.5.0 with no CHANGELOG entry and links to the 17.6.0 section. The original text is unchanged.

Deliberately not in this PR

• content/docs/releases/v17/index.mdx is untouched. Its status blockquote and per-release list may only claim 17.6.0 after the release ships.
• An MDX comment at the top of 17-6.mdx lists the release-time edits:
  1. the publish date;
  2. changesets landed after 748b2407;
  3. the per-package CHANGELOG entry count;
  4. the console-pin outcome;
  5. security(core, plugin-security): an unauthenticated principal never resolves the permission sets bound to the guest anchor; ADR-0090 D9 is declared and seeded but not enforced #21158 if it lands first;
  6. the v17/index.mdx update.

Verification

Run locally on the head commit, with the workspace installed. All of these pass:

• check-issue-citations --base origin/main: every added citation resolves.
• check:doc-anchors, check:role-word, check:docs-audit-scope, check:nul-bytes, check:doc-authoring, check:docs-single-h1, check:docs-redirects, check:docs-locale-catch-all
• check:corpus-claim-drift, check:docs-spec-enumerations, check:published-readme-links
• check:release-notes, check:release-page-status, check:release-index-currency-sync, check:release-body
• check-release-section-coverage (plain and --strict), check-doc-frontmatter, check-docs-nav-label, check-docs-section-name, check-section-landing-index, check-doc-route-spelling --advisory

Both pages compile as MDX with @mdx-js/mdx 3 + remark-gfm, and the three tables parse with no ragged rows.

🤖 Generated with Claude Code

https://claude.ai/code/session_014VGCS11YUtYAiinRcdqQwL

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 4e6dc2338acbc00eaf98bd8269317a779c6b8c02 → packageMentionDocs.