Tuning Oak query indexes — AEM as a Cloud Service

This pattern is executed by the code-assessment runbook — follow ../references/runbook.md for the full flow. This skill supplies the detection + recipe the runbook applies.

Overview

An Oak query only uses an index for the parts the index actually covers. Any WHERE/ORDER BY/fulltext field without a matching, correctly-flagged property definition gets evaluated node-by-node in memory — same for a wrong nodetype or path scope, even if every field is indexed. An uncovered query traverses the repository: slow, traversal-warning-logged, and a load risk at scale. This pattern finds exactly which fields/conditions aren't covered and what index-definition change fixes each one.

Classification — confirm this pattern applies

• A JCR/QueryBuilder query is slow, logs a traversal/TraversingIndex warning, or you're adding/changing a query and need to confirm it's served by an index rather than in-memory filtering or sorting.
• The user asks "is my query indexed", "tune the oak index", "why is this query slow", or a scan flagged a query-construction site (createQuery / PredicateGroup.create).
• Not this pattern: an explicitly unbounded query (p.limit=-1 / setLimit(-1)) — that's unbounded-query; this pattern is about index coverage, not result-set size. (A query can be both — assess index coverage here, bound it there.)

Discovery

Detection is performed by the analyzer (../scripts/analyze.sh), run by the runbook:

bash ../scripts/analyze.sh <workspace-root> --pattern tuning-oak-query-indexes

Match criteria (what the detector flags) — query-construction sites, matched on written names (parse-level, no type resolution). These mirror the Java-API anchors the folded extracting JCR queries guide used, so moving detection from grep to code keeps the same capture:

• any invocation named createQuery — JCR QueryManager.createQuery(...) and AEM QueryBuilder builder.createQuery(...).
• any invocation named getQueryManager — the JCR query entry point (Workspace.getQueryManager()).
• any invocation named findResources or queryResources — the Sling ResourceResolver query APIs, which run a real indexed repository query.
• PredicateGroup.create(...) (a create invocation whose receiver's trailing simple name is PredicateGroup) and new PredicateGroup(...) — QueryBuilder predicate-group construction.

One finding per distinct source line, with the call as the snippet. The detector locates where queries are built; it does not judge coverage — that requires the query text (read the surrounding code) and the index definition(s) (see Resolution contract).

Known limits (out of a parse-level Java detector's reach — use the extracting guide's manual fallback): non-Java queries (XPath/SQL2 in JSP/HTL/config, stored dam:query / Smart-Collection predicate strings, query-string form) and queries hidden behind a non-JCR wrapper method whose body has no direct query call.

Resolution contract

user-supplied — the required input is the Oak/Lucene index definition(s) the query could use. The whole assessment is a comparison of the query against its candidate index definition(s), so:

• Required input — the index definition(s). This skill does not run without them. Accept any of: a FileVault .content.xml under an /apps/.../install package, a JSON export, or the live dump at GET /system/console/status-oak-index-defn.json (admin auth).
• If no index definition is available, DEFER THE WHOLE PATTERN with a single line — e.g. tuning-oak-query-indexes: deferred — index definition not provided. Do not emit a per-query skip for every finding, and do not emit a coverage assessment from the query alone (a guess reads as a finding and gets acted on as one).
• When index definitions ARE supplied, assess each flagged query's coverage and propose the index-definition change where a field is uncovered. fix: guided — never auto-apply: write the corrected definition out for the developer to review and apply.

Recipe

Read recipe.md in full before assessing or proposing any change: the input contract, the Section A/B segregation, the field-construct → required-flag table, the two scoping gotchas, the index-size/storage pitfalls, the common mistakes, the target-index selection, explain verification, and the two-section report shape. Ground truth for every claim is the bundled AI agent indexing guide; build a query inventory from scratch with the bundled extracting JCR queries guide.

Handoff

The skill never commits and never applies an index change to a live instance. See ../references/runbook.md for the full flow and handoff.