# Build Reliable Hybrid Search

Single retrieval methods often miss exact terms or semantic matches. This skill helps combine vector and keyword results with practical fusion and reranking patterns.

## Install

```bash
npx skillstore add sickn33/hybrid-search-implementation
```

## Metadata

- Status: approved
- Slug: sickn33-hybrid-search-implementation
- Skillstore revision: r2
- Version status: missing
- Tree hash: 2dcf2983074ef150e4970ce0f4da071b674d8eafd5bbdfcfc88bd962acf8dbb7
- Author: sickn33
- GitHub username: sickn33
- License: MIT
- Repository: https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/hybrid-search-implementation
- Ref: 88a8e9a07f4c54ab105c1c41b6267c287146b07b
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/sickn33-hybrid-search-implementation
- Manifest: https://skillstore.pages.dev/api/skills/sickn33-hybrid-search-implementation/manifest

## Capabilities

- Explains hybrid search architecture and common fusion methods.
- Provides reciprocal rank fusion and weighted score combination patterns.
- Outlines PostgreSQL retrieval with pgvector and full-text search.
- Shows Elasticsearch patterns for dense vectors, BM25, and reciprocal rank fusion.
- Designs a custom RAG pipeline with retrieval, fusion, filtering, and reranking.

## Use Cases

- Improve RAG Retrieval: Combine semantic recall with exact term matching before generation.
- Modernize Product Search: Add vector retrieval and fusion to an existing keyword search service.
- Plan Retrieval Experiments: Compare fusion strategies, candidate counts, weights, latency, and ranking quality.

## Prompt Templates

### Choose a Fusion Method

```
Explain whether reciprocal rank fusion or weighted score fusion fits my search system. My data is [data], and success means [metric].
```

### Design a Hybrid Search Flow

```
Design a hybrid search flow for [use case] using [vector store] and [keyword engine]. Include candidate retrieval, fusion, and evaluation.
```

### Adapt a Database Implementation

```
Adapt the hybrid search patterns to [database and version]. Use my schema [schema summary] and constraints [constraints]. Address safe filtering and indexing.
```

### Optimize and Evaluate Retrieval

```
Create an experiment plan for my hybrid retriever. Compare [methods], define offline and online metrics, control latency, and diagnose query segments.
```

## Limitations

- The examples require adaptation for each database schema, model, and deployment.
- The skill does not create embeddings or select an embedding model automatically.
- Search quality still requires representative evaluation data and empirical tuning.
- The PostgreSQL metadata filter example needs identifier validation before production use.

## Best Practices

- Evaluate retrieval with representative queries and relevance judgments.
- Log component ranks and scores so ranking failures remain diagnosable.
- Validate filter fields and parameterize every value before executing database queries.

## Anti Patterns

- Do not combine raw scores from different retrievers without normalization.
- Do not tune weights using a few anecdotal queries.
- Do not pass untrusted field names or query structures directly to search backends.

## Security Audit

- Audited at: 2026-07-24T00:44:47.787\+00:00
- Summary: All 32 static findings are false positives caused by keywords, Markdown backticks, headings, or a database fetch method. A separate high-severity SQL injection risk exists because the PostgreSQL example interpolates metadata keys into a query.

## Stats

- Views: 111
- Downloads: 9
- Favorites: 0
- Popularity score: 0