docs: `data-engine.mdx` `EngineQueryOptions` block omits `searchFields`, declared on the schema since #4371

[os-help]

Observation-class docs gap, recorded while implementing #7057 (the cursor / distinct / WriteObservabilityOptions alignment on the same page). Filed unassigned — recording only, no ownership taken. Deliberately NOT fixed in #7057's PR: that card's dispatch scoped the surface to the three ruled divergences, and this is a fourth of a different class (an omission, not a retired key still documented).

The gap

content/docs/kernel/contracts/data-engine.mdx opens its EngineQueryOptions code block with "Defined by EngineQueryOptionsSchema in @objectstack/spec", then lists every member of that schema except searchFields.

searchFields is a real, declared, enforced option — packages/spec/src/data/data-engine.zod.ts (verified on origin/main @ 5087ac6):

  searchFields: z.array(z.string()).optional(),

Its own doc comment records that it was "enforced but undeclared until #4371 (option 2) made the engine reject undeclared option keys" — so it has been a first-class authorable key since that change. The generated reference page already carries it (content/docs/references/data/data-engine.mdx, ## EngineQueryOptions table); only the hand-written contract page is missing it.

Why observation-class, not a defect

Nobody following this page breaks: the omission costs a reader knowledge of one option, it does not make them write something the schema rejects. That is the opposite direction from #7057's ruled divergences, which documented keys the schema now refuses by name. Grading left to triage.

Fix sketch

One line in the code block, alongside search, e.g.

searchFields?: string[]; — fields the search expansion may match against, intersected with the object's declared/derived searchable set (ADR-0061).

Refs: #7057 (the card this was found under), #4371 (the change that made the key declared), ADR-0061.

---

Generated by Claude Code

[os-help]

Claim: PM loop round 7 — domain:spec-tooling seat (#6018)
Session: session_01KJATVrh6V2ysutYUJigh3B
Branch: claude/issue-7170-data-engine-searchfields
Worktree: objectstack-issue-7170
File surface: content/docs/kernel/contracts/data-engine.mdx only (one line in the EngineQueryOptions block). Stop on breach; explain in the report.

This card was filed by this lane's own dev under #7057 and deliberately left out of that PR as a fourth divergence of a different class; triage has since graded it and returned it to pm:queue. Same page as the merged #7173, so the lane has per-path familiarity — which is exactly why the dispatch pins re-reading rather than reusing it.

Premise re-verified live on origin/main before claiming:

• searchFields is still declared — packages/spec/src/data/data-engine.zod.ts:149, searchFields: z.array(z.string()).optional(). The card's premise holds.
• The page is hand-written (not under the merge=os-regen content/docs/references/** tree), so the edit is not regenerated away.
• ⚠️ The page has moved since this card was filed. 8b06bba84 (EngineQueryOptionsSchema.search rejects the bare query string that ADR-0061 D1 calls canonical — so every engine caller that wants it must as any, losing the whole query's checking #7178 → PR fix(spec): EngineQueryOptionsSchema.search accepts the canonical bare query string (#7178) #7249) landed after our docs(kernel): align data-engine contract page with the shipped engine seam (#7057) #7173 and changed EngineQueryOptionsSchema.search to accept the canonical bare query string — i.e. it edited the neighbouring line this addition sits next to. The card's line anchors are from before that commit. Anchor on text, never on line numbers, and re-read the block live; the sibling search prose may already say something the new line must agree with rather than contradict.
• No open PR touches this path.

sonnet — mechanical S-card: one line, in a block whose content is already independently pinned by the generated reference page (content/docs/references/data/data-engine.mdx, ## EngineQueryOptions table), which the dev should diff the hand-written block against rather than trusting the card's suggested wording.

---

Generated by Claude Code