skills.rest
lfuuu
lfuuuCOMMUNITY
Explore all skills by this author →
GitHub

perplexity

Routes web search and research queries to Perplexity Search or Ask with token limits.

Updated Jun 2, 2026
Search Tags:#research#web search#mcp#perplexity#context7#token budget#tool routing
One-click install
npx skills add https://github.com/lfuuu/claude-rules --skill perplexity-lfuuu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: perplexity
Source: https://github.com/lfuuu/claude-rules/tree/main/global-skills/perplexity
Command: npx skills add https://github.com/lfuuu/claude-rules --skill perplexity-lfuuu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Choosing the wrong search tool wastes tokens and bloats context. This Skill defines clear routing rules so Perplexity Search and Ask are used only for general web queries, while library docs, workspace questions, and deep research go to the right specialized tools. ## Core Features & Use Cases - Tool routing chain: Enforces a priority order — Context7 for library docs, Graphite MCP for gt CLI, Nx MCP for workspace questions, then Perplexity Search or Ask, with WebSearch as last resort. - Token-budget defaults: Mandates conservative parameters (max_results: 3, max_tokens_per_page: 512) to prevent context bloat, with explicit criteria for when to raise limits. - Research guardrail: Forbids the costly perplexity_research tool (30-50k tokens) and delegates deep multi-source research to the Researcher agent instead. - Use Case: A user asks "Find postgres migration best practices" — the Skill routes this to Perplexity Search with conservative limits, while a request like "Search for React hooks documentation" is redirected to Context7. ## Quick Start Ask the assistant to search the web for a general topic, for example: "Find the latest best practices for database migrations."

Frequently Asked Questions about perplexity

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I use Perplexity Search without bloating context?▼

Call perplexity_search with max_results set to 3 and max_tokens_per_page set to 512 instead of the default 10 results. Only raise limits to 5 results and 1024 tokens when the user explicitly requests comprehensive results or the first search returned nothing useful.

When should I use Perplexity Ask instead of Perplexity Search?▼

Use Perplexity Ask when you need a conversational explanation synthesized from web sources, such as explaining how postgres advisory locks work. Use Perplexity Search when you need raw results, URLs, guides, or blog articles.

Should I use Perplexity for library or framework documentation?▼

No. Library and framework documentation queries like React hooks or Next.js routing should go to Context7 MCP first. Perplexity is reserved for general web searches that specialized tools do not cover.

Why is the perplexity_research tool forbidden?▼

The perplexity_research tool consumes 30-50k tokens per call, making it extremely costly. Deep multi-source research should be delegated to the Researcher agent via /research, which provides synthesized results with citations.

What is the correct tool selection order for search queries?▼

The priority chain is: Context7 MCP for library docs, Graphite MCP for gt CLI, Nx MCP for workspace questions, then Perplexity Search, Perplexity Ask, the Researcher agent for deep research, and WebSearch only as a last resort.

Related Skills

Explore More Software Engineering Skills→
Shubh2310-developerShubh2310-developer
2

perplexity

Routes generic web search and research queries to Perplexity Search or Ask with token-limited defaults.

Community
Intermediate
sun2443sun2443

search

Search multiple connected MCP sources and synthesize cited results.

Community
Intermediate
ArcReelArcReel
4.3k

research

Investigate questions against primary sources and save findings as cited Markdown notes.

Official
Basic
pedronauckpedronauck
601

perplexity

Routes web search and research queries to Perplexity Search or Ask with token-budget limits.

Community
Intermediate