db-index-hygiene
Skill22 starsChanged 5 months agoMITView on GitHub
Audit index bloat — exact-duplicate indexes, redundant indexes whose key is a prefix of another, and unused indexes that only cost write amplification and storage. Module M12. Feeds the Performance & Scale score.
In this repository36 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- auditSkill
- checklistSkill
- db-antipatternsSkill
- db-concurrencySkill
- db-connection-poolingSkill
- db-constraintsSkill
- db-defaults-generatedSkill
- db-engine-selectionSkill
- db-indexingSkill
- db-keysSkill
- db-migration-safetySkill
- db-multitenancySkill
- db-namingSkill
- db-normalizationSkill
- db-orchestratorSkill
- db-partitioning-shardingSkill
- db-platform-fitSkill
- db-query-patternsSkill
- db-referential-integritySkill
- db-replicas-viewsSkill
- db-security-accessSkill
- db-specialized-fitSkill
- db-storage-bloatSkill
- db-temporal-historySkill
- db-types-precisionSkill
- designSkill
- explainSkill
- fixSkill
- introspectSkill
- migrateSkill
- nextSkill
- scoreSkill
- seedSkill
- stack-detectSkill
- startSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 21
runtime truth: never declared unused from static analysis alone.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
<!-- Copyright (c) 2026 Enrique Rocha (@soyenriquerocha) · tododeia.com · id er-f37b7121 -->db-index-hygiene (M12)
The inverse of M11: M11 adds missing indexes, M12 removes the ones that only cost. Every index slows
writes and consumes storage and cache; duplicates and never-read indexes are pure overhead. Feeds the
Performance & Scale score (axis performance, relational Higiene de índices w16).
What it checks
- Exact duplicates — two indexes with the identical key column list (and same type/predicate); one is redundant.
- Redundant prefix — index on
(a)is fully covered by an existing index on(a, b); the shorter one can usually be dropped (B-tree leftmost-prefix rule). - Unused — an index with no scans over a representative window (Tier-2
idx_scan = 0). This is runtime truth: never declared unused from static analysis alone.
Score / axis
Feeds performance only (relational Higiene de índices w16; folds into Indexación in NoSQL profiles).
Tier-0 (static)
Parse declared indexes and detect exact duplicates and leftmost-prefix redundancy from the DDL alone
(established for duplicates — purely structural). Unused-index detection is runtime-only and
emits status: needs_api at Tier-0 (never a silent pass) with a nudge to Tier-2.
Tier-1/2 (verification query, Postgres)
Duplicate/redundant (Tier-1 catalog):
SELECT indrelid::regclass AS tbl, array_agg(indexrelid::regclass) AS dupes
FROM pg_index GROUP BY indrelid, indkey, indclass, indpred
HAVING count(*) > 1;Unused (Tier-2, requires sustained stats):
SELECT relname AS tbl, indexrelname AS idx, idx_scan, pg_size_pretty(pg_relation_size(indexrelid)) AS sz
FROM pg_stat_user_indexes WHERE idx_scan = 0 ORDER BY pg_relation_size(indexrelid) DESC;Method index_check / query_stat. The Tier-2 result confirms M12.*.unused as established;
without a representative window it stays directional, and if stats are needed to decide → needs_api.
Findings
Emit findings per schema/finding.schema.json. Examples:
M12.users.duplicate_email_index— two indexes on the same key (severity:2,warn, axisperformance, confidenceestablishedfrom DDL,fixable: proposed—DROP INDEX).M12.orders.redundant_prefix_index—(status)covered by(status, created_at)(severity:1,warn,established/directional,fixable: proposed).M12.events.unused_index— zero scans over the window (severity:2,warn,establishedTier-2 /needs_apiat Tier-0,fixable: proposed).
Each finding: evidence.observed quotes the index DDL or the catalog/stats row verbatim;
verification.reproduce is the query above referencing $DATABASE_URL; expected_impact is banded +
confidence-tagged (no naked %).
Honesty
- An unused index over a short window may be hit by month-end/quarterly jobs — always require a
representative window and mark short-window verdicts
directional, notestablished. - A unique/constraint-backing index is not droppable even if "unused" for scans — it enforces integrity; never recommend dropping it.
- Drops are
proposed(neverauto): removal is not trivially reversible at scale and intent may be deliberate. SuggestDROP INDEX CONCURRENTLYin the fix preview.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

