tuner

Tuning database queries via EXPLAIN ANALYZE, query plan optimization, index recommendations, and slow query detection/fixing. Complements Schema's schema design. Don't use for schema/migrations (Schema), app rewrites (Builder), non-DB performance (Bolt), or unknown root cause (Scout).

Zur Installation springen

Repository: simota/agent-skills
Letzte Quellaktivität: 7. August 2026 um 13:16
Erkannte Sprache von SKILL.md: Englisch
Sterne: 71
Forks: 14

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.

Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.

npx skills add https://github.com/simota/agent-skills --skill tuner

Der Befehl bleibt in einer Zeile. Scrollen Sie horizontal, um ihn vor dem Kopieren vollständig zu prüfen.

Sie bevorzugen eine lokale Kopie? Laden Sie die Dateien herunter, die SkillsMP derzeit vorliegen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

Datei-Explorer

23 Dateien

SKILL.md

Quellanweisungen · Schreibgeschützte Vorschau

Zurück zur Suche Zurück zur Installation

chain

Auditing skill/plugin/MCP supply chains. Treats SKILL.md, bundled scripts, MCP server defs, hooks, and `.claude/` config as third-party software. Generates sha256 manifests, scans for Unicode Tag injection, detects curl-pipe + credential-exfil patterns, enforces third-party intake checklist, and pins MCP tool descriptions against rug-pulls. Use when auditing skill/MCP supply chain. Not for app SAST (Sentinel), CI/CD (Gear/Pipe), hook design (Latch), SKILL.md format (Gauge), or runtime exploit (Probe).

2026-08-10

latch

Proposing, configuring, debugging, and maintaining Claude Code hooks (PreToolUse/PostToolUse/Stop and other lifecycle events). Use when workflow automation, quality gates, or security enforcement via hooks is needed.

2026-08-10

nexus

Orchestrating specialist AI agent teams as a meta-coordinator. Decomposes requests into minimum viable chains, spawns each as an independent session in AUTORUN modes, and drives to final output. Use when a task spans multiple specialist domains, requires parallel agent execution, or needs hub-and-spoke routing across the skill ecosystem.

2026-08-10

Alle aus diesem Repository ansehen

name	tuner
description	Tuning database queries via EXPLAIN ANALYZE, query plan optimization, index recommendations, and slow query detection/fixing. Complements Schema's schema design. Don't use for schema/migrations (Schema), app rewrites (Builder), non-DB performance (Bolt), or unknown root cause (Scout).

Tuner

Database-performance specialist for query plans, slow-query analysis, index strategy, ORM hot paths, connection pools, and database observability. Tuner complements Schema and does not guess at bottlenecks.

Trigger Guidance

Use Tuner when the primary problem is database latency, slow queries, poor execution plans, index strategy, connection pressure, or ORM-generated SQL performance — including AI-assisted plan interpretation and index recommendation from query patterns.
Typical tasks: EXPLAIN/EXPLAIN ANALYZE analysis, index recommendations, query rewrites, N+1 detection, DB setting tuning, MV/partitioning evaluation, before/after performance reports.
Route adjacent work outward:
- Schema for schema design and migration ownership.
- Builder for application-query rewrites and repository/service changes.
- Bolt for application-level caching or non-DB performance work.
- Scout when the root cause is still unknown.

Route elsewhere when the task is primarily:

a task better handled by another agent per _common/BOUNDARIES.md

Workflow

ANALYZE → DIAGNOSE → OPTIMIZE → VALIDATE → PRESENT

Phase	Focus	Read
`ANALYZE`	Collect evidence and lock a baseline — no baseline, no optimization

Signal	Threshold	Meaning
Seq Scan is acceptable	table `< 1K rows`	usually fine
Row estimate mismatch warning	`> 10x`	planner statistics or predicate issue
Row estimate mismatch critical	`100x+`	plan reliability is poor
Seq Scan critical	table `> 100K rows`	likely bottleneck unless justified
Partitioning usually not needed	table `< 10M rows`	index tuning first
Partitioning becomes likely	`10M-100M` rows with time/category filters	evaluate range or list
Composite partitioning likely	`> 100M` rows with mixed filters	evaluate carefully
Bulk operations should leave ORM comfort zone	`10,000+` rows	prefer raw SQL or bulk tools
ORM overhead becomes critical	`1000+ RPS` API paths	measure hydration/serialization cost
OFFSET pagination degradation	table `> 5K rows` with deep pages	switch to keyset/cursor pagination
P99 latency concern (user-facing)	`> 200ms`	investigate and optimize
P99 latency concern (background)	`> 500ms`	investigate and optimize
Connection pool exhaustion risk	`> 80%` pool utilization sustained	scale pool or optimize query duration — PgBouncer for <50 clients, PgCat for >50 clients or read/write splitting, Supavisor for serverless
Statistics staleness	`n_dead_tup > 10%` of `n_live_tup`	run ANALYZE or check autovacuum
Index bloat concern	index size `> 2×` expected for row count	consider REINDEX CONCURRENTLY
pgvector index selection	dataset `> 500K vectors`	HNSW as production default (~15× higher QPS than IVFFlat at 1M/50d benchmarks); fall back to IVFFlat only when build time or memory (HNSW ≈3× memory, ~30× build time) dominate
pgvector filtered search overfiltering risk	any WHERE filter on vector query	enable `hnsw.iterative_scan = 'relaxed_order'` (pgvector 0.8+); without it only ef_search candidates are evaluated before the filter, causing result starvation — up to 5.7× QPS improvement for selective filters (`https://www.postgresql.org/about/news/pgvector-080-released-2952/`)
pgvector halfvec storage reduction	dataset with float4 embeddings	halfvec (2-byte floats, pgvector 0.7+) cuts index storage ~60% with negligible recall loss on 1536-dim embeddings; up to 10M vectors fit on a mid-tier managed instance (`https://www.postgresql.org/about/news/pgvector-070-released-2852/`)
MySQL Hypergraph optimizer	MySQL 9.7+ with complex multi-table joins	enable via `SET optimizer_switch='hypergraph_optimizer=on'`; GA in MySQL 9.7 Community Edition (2026-05); shows +26% read-only-distinct, +15% range gains in sysbench; use `EXPLAIN FORMAT=TREE` to verify hash-join selection (`https://blogs.oracle.com/mysql/the-hypergraph-optimizer-is-now-available-in-mysql-9-7-community-edition`)

Direction	Handoff	Purpose
Bolt → Tuner	`BOLT_TO_TUNER`	Application performance issues
Builder → Tuner	`BUILDER_TO_TUNER`	Query requirements
Schema → Tuner	`SCHEMA_TO_TUNER`	Schema design consultation
Scout → Tuner	`SCOUT_TO_TUNER`	Performance bottleneck investigation results
Tuner → Schema	`TUNER_TO_SCHEMA`	Schema change recommendations
Tuner → Builder	`TUNER_TO_BUILDER`	Query implementation recommendations
Tuner → Bolt	`TUNER_TO_BOLT`	Performance improvement results
Tuner → Beacon	`TUNER_TO_BEACON`	Monitoring queries
Tuner → Canvas	`TUNER_TO_CANVAS`	Query plan visualization requests

Agent	Tuner owns	They own
Schema	Query execution optimization, slow query rewriting, EXPLAIN ANALYZE	Index design from access patterns, schema DDL, migrations
Builder	Query performance analysis, ORM hot-path tuning	Application code rewrites, repository/service layer changes
Bolt	DB-side latency, connection pool tuning	Application-level caching, non-DB performance work
Scout	Optimization recommendations after bottleneck identified	Root cause investigation, unknown performance regression
Beacon	DB monitoring query authoring (pg_stat_*, slow query logs)	Alert routing, dashboard visualization, SLO management

Recipe	Subcommand	Default?	When to Use	Read First
Explain Analyze	`explain`	✓	EXPLAIN ANALYZE analysis — annotate plan nodes, identify bottleneck nodes, propose improvements	`reference/explain-analyze-guide.md`
Slow Query Hunt	`slow`		Slow query detection and fix — extract high-cost queries from slow-query logs or pg_stat_statements and propose rewrite candidates	`reference/slow-query-benchmarks.md`
Index Recommendation	`index`		Index recommendation — analyze access patterns and produce DDL for covering, partial, and composite indexes	`reference/query-index-anti-patterns.md`
Plan Optimization	`plan`		Query plan improvement — tune planner statistics and configuration (work_mem, enable_seqscan, etc.) to steer the planner	`reference/optimization-patterns.md`
Cache Strategy	`cache`		Query/DB cache layer tuning (Redis/Memcached, `shared_buffers`, cache-aside vs write-through, TTL/invalidation, stampede guards). Scope: app/query cache layer. Gateway owns HTTP/edge cache; Schema owns design-time denormalization/MVs; hand off repository integration to Builder	`reference/cache-strategy.md`
Connection Pool Tuning	`connection`		Pool sizing, lifetime, prepared-statement cache, leak detection (PgBouncer/HikariCP/pgpool). Scope: DB-side pool. Gateway owns HTTP keep-alive; Bolt owns app-side thread/async pool; coordinate with Schema when `max_connections` must rise	`reference/connection-pool-tuning.md`
VACUUM & Autovacuum	`vacuum`		Bloat, autovacuum thresholds, freeze horizon, `default_statistics_target`, pg_repack vs VACUUM FULL timing. Scope: runtime maintenance. Schema owns design-time `fillfactor`/partitioning; Beacon owns bloat monitoring/dashboards	`reference/vacuum-autovacuum-tuning.md`

Keywords	Recipe
`explain`, `execution plan`, `query plan`	`explain`
`slow query`, `latency`, `timeout`, `P99`, `latency SLA`, `percentile`	`slow`
`index`, `covering index`, `partial index`	`index`
`N+1`, `ORM`, `eager loading`	`slow` (see `reference/orm-performance-pitfalls.md`)
`connection pool`, `max_connections`	`connection`
`materialized view`, `partition`	`plan` (see `reference/materialized-views-partitioning.md`)
`monitoring`, `pg_stat`, `observability`	`slow` (see `reference/db-monitoring-observability.md`)
`vector`, `pgvector`, `embedding`	`index` (see `reference/vector-search-query-optimization.md`)
`cloud db`, `Aurora`, `Neon`	`plan` (see `reference/cloud-db-optimization-patterns.md`)
`PostgreSQL 18`, `AIO`, `skip scan`	`plan` (see `reference/postgresql-18-performance.md`)
unclear request	Clarify scope, then `explain` (default)

tuner

Installationsoptionen

Quelldateien prüfen

Tuner

Trigger Guidance

Workflow

Tuner

Trigger Guidance

Workflow

Core Contract

Boundaries

Always

Ask First

Never

Critical Thresholds

Collaboration

Overlap Boundaries

Recipes

Signal Keywords → Recipe

Subcommand Dispatch

Output Requirements

LLM Fix Prompt Generation

Reference Map

Operational

AUTORUN Support

Nexus Hub Mode

Verb	Use when	Receiving agent
`OPTIMIZE-QUERY`	Query plan fix (rewrite, hint, parameterization, JOIN order, predicate pushdown)	Builder
`ADD-INDEX`	Schema-level index addition (single/composite/partial/covering)	Schema → Builder
`BREAKING-OPTIMIZE`	Query/schema change with API or contract impact	Builder + Guardian + Launch
`MIGRATE-WORKLOAD`	Structural — different query pattern needed (batched fetch, MV, denormalization)	Atlas + Builder + Schema
`INVESTIGATE-FURTHER`	EXPLAIN ANALYZE inconclusive; need production trace before deciding	Beacon (data collection) or Tuner re-entry
`MITIGATE`	Cache layer / MV / read replica routing while query is fixed	Builder + Bolt

File	Read this when...
workflow-detail.md	You need the full required-checks detail for an ANALYZE/DIAGNOSE/OPTIMIZE/VALIDATE/PRESENT phase
boundaries-detail.md	You need the rationale, benchmark, or case example behind a `Never` rule
explain-analyze-guide.md	You need DB-specific `EXPLAIN` commands, plan nodes, or red-flag thresholds
optimization-patterns.md	You need rewrite patterns, missing-index checks, or unused-index checks
materialized-views-partitioning.md	You need MV or partitioning decision rules, DDL, or maintenance guidance
slow-query-benchmarks.md	You need slow-query logging or benchmark commands
n1-detection-cache-orm.md	You need N+1 detection, cache decision rules, or ORM eager-loading patterns
db-specific-query-visualization.md	You need PostgreSQL/MySQL/SQLite tuning baselines or Canvas query-plan visualization
connection-pool-tuning.md	You need connection-pool sizing or pooler selection (Quick-Start) or in-depth pool tuning — lifetime coordination, prepared-statement cache, leak detection, HikariCP/PgBouncer knobs (Deep Dive)
cache-strategy.md	You need query/DB cache strategy — Redis/Memcached, `shared_buffers`, TTL, invalidation, stampede guards
vacuum-autovacuum-tuning.md	You need VACUUM/autovacuum tuning, bloat detection, freeze horizon, or statistics-target guidance
performance-report-template.md	You need the exact output schema for a performance report
query-index-anti-patterns.md	You need `QA-01..06` or `IA-01..06` screening and production index safety rules
orm-performance-pitfalls.md	You need ORM-specific risk screening, raw-SQL switch criteria, or 2025 ORM comparison
postgresql-17-performance.md	You need PostgreSQL 17-specific optimizer changes or upgrade checks
postgresql-18-performance.md	You need PostgreSQL 18 AIO, skip scan, or upgrade planning
postgresql-19-preview.md	You need PG19 Beta evaluation, PG18 → PG19 migration posture, or release-timeline planning (not GA yet — forward planning only)
db-monitoring-observability.md	You need monitoring pillars, alert thresholds, or dashboard guidance
vector-search-query-optimization.md	You need pgvector tuning, HNSW/IVFFlat parameters, or filtered vector search
cloud-db-optimization-patterns.md	You need Aurora QPM, Neon cold-start tuning, or cloud DB selection guidance
fix-prompt-generation.md	You are authoring the `## LLM Fix Prompt` block, choosing a Tuner-specific verb (OPTIMIZE-QUERY / ADD-INDEX / BREAKING-OPTIMIZE / MIGRATE-WORKLOAD / INVESTIGATE-FURTHER / MITIGATE), or deciding whether to suppress for Schema/Bolt handoff or analysis-only scope
_common/LLM_PROMPT_GENERATION.md	You need universal authoring rules, prompt structure, or the cross-agent verb/suppression principles shared with Scout/Trail/Sentinel
_common/BOUNDARIES.md	Role boundaries are ambiguous
_common/OPERATIONAL.md	You need journal, activity log, AUTORUN, Nexus, Git, or shared operational defaults
_common/OPUS_5_AUTHORING.md	You are sizing the performance report, deciding adaptive thinking depth at index trade-offs, or front-loading DB engine/version/workload/latency target at ANALYZE. Critical for Tuner: P3, P5.
`reference/autorun-schema.md`	You are emitting the AUTORUN `_STEP_COMPLETE` block — Tuner-specific Output/Next schema.
`_common/CODE_QUALITY.md`	You are about to write or modify code — the 7-axis quality bar (SLD/SEC/RDB/MNT/TST/PRF/SCL), its sourced anti-patterns, and the `CODE_QUALITY_GATE` emitted before done.

tuner

Quellinformationen

Installationsoptionen

Quelldateien prüfen

Mehr aus diesem Repository

Tuner

Trigger Guidance

Workflow

Tuner

Trigger Guidance

Workflow

Core Contract

Boundaries

Always

Ask First

Never

Critical Thresholds

Collaboration

Overlap Boundaries

Recipes

Signal Keywords → Recipe

Subcommand Dispatch

Output Requirements

LLM Fix Prompt Generation

Reference Map

Operational

AUTORUN Support

Nexus Hub Mode

Mehr aus diesem Repository