niche-library-research

---
name: niche-library-research
description: "Dispatch a swarm of research agents to learn a library's API before writing code. Use this skill BEFORE writing any code that uses a library you're not confident about — especially niche libraries like ELK.js, obscure layout engines, specialized rendering tools, or any library with fewer than 5,000 GitHub stars. Also use when: code you wrote with a library isn't working and you're not sure why, the user says 'research this first', 'send agents', 'look up the docs', or you catch yourself guessing at API parameters. If you're about to write configuration for a library and you're not 100% certain of the exact option names and values, use this skill. LLMs hallucinate at 2.4x the rate on low-frequency APIs — this skill prevents that."
---

# Niche Library Research

When you hit a knowledge wall — a library whose API you're not confident about — stop coding and dispatch research agents. This is faster than guessing, iterating, and debugging hallucinated APIs.

## When to Use

- Before writing code with any library you haven't used successfully in this session
- When a library has complex configuration (many options, Java-style naming, sparse docs)
- When your first attempt with a library didn't work
- When the user directs you to research before coding
- When you're about to guess at parameter names or values

## The Pattern

Dispatch 2-3 agents in parallel, each targeting a different source:

### Agent 1: Official Documentation
```
Search the official documentation for [library].
I need to understand [specific feature/API].
Find: exact function signatures, configuration options with their types and defaults,
and any gotchas or common mistakes.
URL: [official docs URL if known]
Give me concrete answers with code examples, not theory.
```

### Agent 2: GitHub Issues & Discussions
```
Search GitHub issues and discussions for [library repo].
I'm trying to [specific task].
Find: real-world configurations that work, common pitfalls people hit,
workarounds for known issues, and any undocumented behavior.
Search terms: [relevant keywords]
Give me concrete configs and code, not theory.
```

### Agent 3: Real-World Usage
```
Search for real-world projects, blog posts, and Stack Overflow answers
that use [library] for [specific use case].
Find: complete working configurations, not fragments.
For each example, extract the exact config/code they used and explain why it works.
```

## After Research Returns

1. Synthesize findings into a concrete plan before writing code
2. Cite which source each piece of information came from
3. Flag any contradictions between sources
4. If the research didn't answer your question, say so — don't fill gaps with guesses

## Key Principle

The cost of a 3-minute agent dispatch is always less than the cost of guessing wrong, claiming it works, getting caught, and having to research anyway. Research first, code second.