hybrid-search-implementation Skill (wshobson) for Claude: What It Does and How to Install It

Helps Claude combine vector and keyword search for better retrieval, with fusion methods and Postgres, Elasticsearch and pipeline templates.

Installs: 11K; on skills.sh
Stars: 42.1K; on GitHub
Files: 1; besides SKILL.md
Updated: Oct 2026; last repo commit

This is a retrieval guide for Claude, from wshobson's agents collection. SKILL.md shows the hybrid architecture (vector and keyword candidates fused into results), a table of four fusion methods (RRF, linear, cross-encoder and cascade) and do and don't lists. references/details.md holds four Python templates: Reciprocal Rank Fusion, PostgreSQL hybrid search, Elasticsearch hybrid search and a custom hybrid RAG pipeline.

It loads when you build RAG systems or search engines, or when neither vector nor keyword search alone gives enough recall. It bundles no scripts. It needs your own database or search index.

Good for

Engineers building RAG systems that miss exact terms like names and codes
Teams adding keyword search next to an existing vector index
Developers tuning fusion weights and reranking

Install hybrid-search-implementation

Claude Code

Copy this prompt.

Install the hybrid-search-implementation skill in Claude Code. Run this command:

/plugin marketplace add wshobson/agents
/plugin install llm-application-dev@claude-code-workflows

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Pick one route. The plugin routes install the whole llm-application-dev plugin: 3 agents, 3 commands and 8 skills (counted from the plugin folder in the repo tree, which has no hooks or MCP config files). The skills-only routes install just this skill, with no agents or commands. The skill needs a search backend of your own, such as PostgreSQL or Elasticsearch.

From the project README. Source.

What's in the skill

When it loads

Combine vector and keyword search for improved retrieval. Use when implementing RAG systems, building search engines, or when neither approach alone provides sufficient recall.

Sections in SKILL.md

When to Use This Skill
Core Concepts
1. Hybrid Search Architecture
2. Fusion Methods
Templates and detailed worked examples
Best Practices
Do's
Don'ts

Key facts

Author: wshobson (repo owner)
License: MIT
Installs on skills.sh: 10,995 (Nov 2, 2026)
GitHub stars: 42,081 for the whole wshobson/agents repo (Nov 2, 2026)
Last repo commit: Oct 30, 2026
What it includes: 2 files including SKILL.md: references/details.md holds the 4 templates (no scripts)
Needs: A search backend of your own. The templates use PostgreSQL with asyncpg, Elasticsearch and sentence-transformers
Part of: The llm-application-dev plugin in wshobson/agents. Counted from the plugin folder: 3 agents, 3 commands and 8 skills
Works in: Claude Code, Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot and Pi, per the README. Skills-only installers ask which tool you use

What it includes

Vector and keyword candidates merged by a fusion step
A query runs through both searches, and a fusion step combines the candidates into results.
Try: “Add keyword search next to our vector search and fuse the results.”
Read · Hybrid search architecture
RRF, linear, cross-encoder and cascade
RRF is listed as general purpose, linear as tunable, cross-encoder as highest quality and cascade as efficient.
Read · Fusion methods
Python code that merges ranked lists
Combines result lists from several searches by rank.
Write · Template 1: Reciprocal Rank Fusion
Vector table, indexes and a cross-encoder rerank option
Uses asyncpg, CREATE EXTENSION vector, an embedding index and a full-text index.
Write · Template 2: PostgreSQL hybrid search
Python client for vector plus keyword queries
Uses the elasticsearch Python package.
Write · Template 3: Elasticsearch hybrid search
A HybridRAGPipeline class with pluggable stores and reranker
Runs vector and keyword searches in parallel, fuses them and optionally reranks.
Write · Template 4: custom hybrid RAG pipeline
Tune weights on your data, log both scores, A/B test
Also: use RRF for simplicity, add reranking, keep keyword search and handle empty or single-word queries.
Read · Do's and don'ts

From the official docs

Permissions

Low risk

The main risk is that generated schema changes and queries run against a real database. Try them on a development database first.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

The Postgres template creates an extension, a table and indexes
details.md has CREATE EXTENSION IF NOT EXISTS vector, CREATE TABLE IF NOT EXISTS documents and index statements. Run them on a development database and review them first. github.com
The plugin route installs more than this skill
The llm-application-dev plugin has 8 skills, 3 agents and 3 commands. The repo tree shows no hooks or MCP config files in its folder. The skills-only routes install only this skill. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only mode. Use version control.
Talks to	Connections to the PostgreSQL or Elasticsearch instance you configure
Scopes	Claude writes search and retrieval code in your project when you ask The PostgreSQL template contains CREATE EXTENSION vector, CREATE TABLE and CREATE INDEX statements The templates connect to the Postgres or Elasticsearch instance you configure

Questions

What does the hybrid-search-implementation skill do?

It gives Claude patterns for mixing vector search and keyword search. Claude uses it to write fusion code and set up Postgres or Elasticsearch so exact terms and meaning both count.

How do I install hybrid-search-implementation?

Use the install box at the top of this page. It lists the plugin route and two skills-only commands: npx skills add wshobson/agents --skill hybrid-search-implementation and gh skill install wshobson/agents hybrid-search-implementation (GitHub CLI 2.90 or later). The skills-only commands skip the plugin's agents and commands.

Which fusion method should I start with?

SKILL.md lists RRF (Reciprocal Rank Fusion) as general purpose, and its do's say to use RRF for simplicity because it works well without tuning. It recommends tuning weights on your own data.

Which databases do the templates use?

PostgreSQL with the vector extension, and Elasticsearch. Template 4 takes any vector store and keyword store you pass in.

Does it need an embedding API key?

SKILL.md and details.md name none. The templates take an embedding or an embedder object that you provide, so any key depends on the model you choose.

Is it safe?

Installing it adds a guide and one reference file, with no scripts. The Postgres template contains CREATE EXTENSION, CREATE TABLE and CREATE INDEX statements, so run them on a development database first and review what Claude generates before it touches production data.

Is this skill from Anthropic?

No. It is in wshobson/agents, a community collection. The skill sits in the llm-application-dev plugin there.

Should I install the whole llm-application-dev plugin?

Only if you want the rest. The README documents a plugin route for Claude Code and Codex. The plugin folder holds 8 skills, 3 agents and 3 commands, and the repo tree shows no hooks or MCP config files in it.

Information

updated Nov 2, 2026

Repository: wshobson/agents
Folder: plugins/llm-application-dev/skills/hybrid-search-implementation

wshobson/agents skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 4

hybrid-search-implementation SKILL.md · github.com
wshobson/agents README · github.com
llm-application-dev plugin.json · github.com
hybrid-search-implementation details.md · github.com