ai-rag-pipeline — Claude Code skill: RAG with web search and LLM
ai-rag-pipeline — a skill for building RAG pipelines: web search supplies the facts, a model writes the cited answer. Handy for research agents and claim checking.
no license 240 installs ★ 31 updated 2026-08-31
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a claude-codeWhat ai-rag-pipeline does
The skill lays out how to assemble retrieval pipelines from hosted apps on inference.sh. Search brings back fresh results, the model receives them inside the prompt and answers with sources attached. You get a few shapes to pick from: one search plus one answer, two search engines merged into a single analysis, or URL content pulled and condensed.
Commands run as belt app run against tavily/search-assistant, exa/search, exa/answer, tavily/extract, and models routed through openrouter/ such as GPT-4o and Gemini 2.5 Pro. The agent stores search output in a shell variable and pastes it into the next prompt. Claim checking has its own flow: gather evidence first, then return a verdict alongside supporting and contradicting material.
You need the belt CLI (npx skills add belt-sh/cli) plus a session from belt login; belt app run fails without it. The skill's permission is scoped to Bash(belt *), so it stays out of the rest of your shell. Long search dumps are meant to be summarised first — otherwise the model's context fills up fast.
What it helps with
- answer a question from live search results with sources attached
- fact-check a claim against collected evidence and get a verdict
- merge Tavily and Exa output into one analysis that flags contradictions
- extract readable text from a set of URLs and boil it down
- produce a topic report with stats and expert commentary
Example prompts for the agent
- Run a RAG pass on quantum computing: search with
tavily/search-assistant, then haveopenrouter/write a summary with citations - Fact-check a claim about generative AI in creative work — gather research first, then give me a verdict and sources
- Pull the content from these URLs with
tavily/extractand list the key findings
How to install ai-rag-pipeline
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a claude-code2Manually: put the files in place
~/.claude/skills/ai-rag-pipeline/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/ai-rag-pipeline/
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a codex2Manually: put the files in place
.agents/skills/ai-rag-pipeline/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a cursor2Manually: put the files in place
.cursor/skills/ai-rag-pipeline/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install 101-skills/superpowers tools/llm/ai-rag-pipeline2Manually: put the files in place
.github/skills/ai-rag-pipeline/SKILL.md1Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a gemini-cli2Platform CLI
gemini skills install https://github.com/101-skills/superpowers --path tools/llm/ai-rag-pipelineManually: put the files in place
.devin/skills/ai-rag-pipeline/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about ai-rag-pipeline
How do I install ai-rag-pipeline in Claude Code?
Run npx skills add 101-skills/superpowers --skill ai-rag-pipeline -a claude-code. The skill file ends up at ~/.claude/skills/ai-rag-pipeline/SKILL.md.
What does ai-rag-pipeline do?
It gives the agent a working recipe for pairing web search with a language model: results go into the prompt, the answer comes back cited. Separate flows cover claim checking and report writing.
What else do I need for it to work?
The belt CLI, installed with npx skills add belt-sh/cli, and a login through belt login. Everything the skill calls goes through belt app run.