Agent skill

RAG

by automateyournetwork in automateyournetwork/netclaw

Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base.

Apache-2.0Auto-check passedAI & LLM Engineering

Install RAG

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .claude/skills && cp -r skills-src/workspace/skills/rag .claude/skills/rag && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Paste into Claude Code

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .claude/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

Claude Code copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

$skill-installer

$skill-installer install https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag

Type this inside Codex. $skill-installer <name> installs a curated skill from openai/skills. The installer writes to $CODEX_HOME/skills (default ~/.codex/skills). Restart Codex if the skill does not show up.

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a codex

Project install goes to .agents/skills/; add -g for ~/.codex/skills/.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag --agent codex

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .agents/skills && cp -r skills-src/workspace/skills/rag .agents/skills/rag && rm -rf skills-src

Use ~/.agents/skills/ instead of .agents/skills for a personal install.

Codex skills documentation · loads skills from .agents/skills/

Paste into Codex

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .agents/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

Codex copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a cursor

Project install goes to .agents/skills/; add -g for ~/.cursor/skills/.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag --agent cursor

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .cursor/skills && cp -r skills-src/workspace/skills/rag .cursor/skills/rag && rm -rf skills-src

Use ~/.cursor/skills/ instead of .cursor/skills for a personal install.

Cursor skills documentation · loads skills from .cursor/skills/, .agents/skills/, .claude/skills/, .codex/skills/

Paste into Cursor

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .cursor/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

Cursor copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

gemini skills install

$ gemini skills install https://github.com/automateyournetwork/netclaw.git --path workspace/skills/rag

--scope user (default) or --scope workspace; --path is the subfolder of the repo that holds the skill; --consent skips the security confirmation prompt.

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a gemini-cli

Project install goes to .agents/skills/; add -g for ~/.gemini/skills/.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag --agent gemini-cli

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .gemini/skills && cp -r skills-src/workspace/skills/rag .gemini/skills/rag && rm -rf skills-src

Use ~/.gemini/skills/ instead of .gemini/skills for a personal install, then run /skills reload.

Gemini CLI skills documentation · loads skills from .gemini/skills/, .agents/skills/

Paste into Gemini CLI

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .gemini/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

Gemini CLI copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag

Installs for Copilot at project scope by default; add --scope user for a personal install. Preview a skill first with gh skill preview. Needs GitHub CLI 2.90.0 or later (public preview).

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a github-copilot

Project install goes to .agents/skills/; add -g for ~/.copilot/skills/.

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .github/skills && cp -r skills-src/workspace/skills/rag .github/skills/rag && rm -rf skills-src

Use ~/.copilot/skills/ instead of .github/skills for a personal install. Commit .github/skills so cloud agent and code review can use it.

GitHub Copilot skills documentation · loads skills from .github/skills/, .claude/skills/, .agents/skills/

Paste into GitHub Copilot

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .github/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

GitHub Copilot copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

skills CLI

$ npx skills add automateyournetwork/netclaw --skill rag -a opencode

OpenCode documents no install command of its own. Project install goes to .agents/skills/; add -g for ~/.config/opencode/skills/.

GitHub CLI

$ gh skill install automateyournetwork/netclaw rag --agent opencode

Project scope by default (.agents/skills/); add --scope user for a personal install.

Manual copy

$ git clone --depth 1 https://github.com/automateyournetwork/netclaw.git skills-src && mkdir -p .opencode/skills && cp -r skills-src/workspace/skills/rag .opencode/skills/rag && rm -rf skills-src

Use ~/.config/opencode/skills/ instead of .opencode/skills for a personal install.

OpenCode skills documentation · loads skills from .opencode/skills/, .claude/skills/, .agents/skills/

Paste into OpenCode

Install the "rag" agent skill from https://github.com/automateyournetwork/netclaw/tree/main/workspace/skills/rag into .opencode/skills/rag/ in this project. Copy the whole folder (SKILL.md and every file beside it), keep the folder name "rag", then confirm the skill loads.

OpenCode copies the folder itself, the same result as the manual copy. Check what it changed before you commit it.

Facts

Skill name: rag
GitHub stars: 676
Token cost: ~2.8k tokens
SKILL.md length: 1,369 words
Files: 1
Skills in repo: 120
Repo updated: Oct 5, 2026
First seen: Oct 7, 2026
Licence: Apache-2.0

At a glance

Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base.

Works in 5 steps: Route first (Adaptive RAG) → Rewrite and decompose → Critique after every retrieval (Self-RAG) → …
Document questions
SKILL.md covers Overview, MCP Tools, The Four Knowledge Sources… and Workflow: Slack Document Upload, plus 6 more sections
Calls python3

What it does

RAG is an agent skill from automateyournetwork/netclaw. Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base. Use for document questions, corpus management, or explicitly requested snapshots.

Its SKILL.md is about 2.8k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in AI & LLM Engineering, covering Retrieval-augmented generation and Knowledge bases. The repository describes itself as: An AI agent that claws through your network. The licence is Apache-2.0.

When your agent uses it

Document questions
Corpus management
Explicitly requested snapshots

Example prompts

“/rag”

Requirements

Python 3

Workflow steps

5 steps, taken from the step headings in SKILL.md.

Route first (Adaptive RAG)
Rewrite and decompose
Critique after every retrieval (Self-RAG)
Respect the iteration budget — 3 rounds per sub-query
Honest miss

What it can do on your machine

Read from SKILL.md and the folder at commit 95bb17e. It shows what the files ask for, not the result of running them.

Tool permissions
Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.
From allowed-tools in the SKILL.md frontmatter.
Runs code
Shell commands in SKILL.md call:
- python3
From the folder's file list and the shell code blocks in SKILL.md.
Network
No URLs in SKILL.md.
From URLs in SKILL.md, links to its own repository left out.
Credentials
Names no API keys, tokens, secrets or passwords.
From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

RAG loads about 2.8k tokens when it runs. Until then it costs about 53 tokens; SKILL.md has 1,369 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it: ~53
When it runs · the whole SKILL.md, loaded when a task matches: ~2.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from automateyournetwork/netclaw at commit 95bb17e, republished under its Apache-2.0 licence (© automateyournetwork). 1,369 words, ~2,771 tokens.

Download SKILL.mdSave it as .claude/skills/rag/SKILL.md (or your agent's skills folder).

name: rag
description: Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base. Use for document questions, corpus management, or explicitly requested snapshots.

Skill: RAG Knowledge Base

Purpose: Give NetClaw a fully offline, user-curated document knowledge base — vendor guides, standards (RFC/IEEE/vendor), customer design documents, install guides — with agentic retrieval, mandatory citations, and opt-in point-in-time snapshots.

Overview

Users teach NetClaw by uploading documents (Slack attachment, HUD Knowledge panel, or URL). Documents are parsed, chunked structure-aware, embedded locally, and stored at ~/.openclaw/rag/. Retrieval is a tool NetClaw invokes on its own judgment — iteratively, with self-critique — never a fixed pipeline.

This is NOT memory. The knowledge base holds only what users deliberately put into it. NetClaw's own experience (facts, session summaries, decisions, entity graphs) lives in the Memory MCP (memory_* tools, ~/.openclaw/memory/). Neither store writes into the other.

MCP Tools

Tool	WHEN to use
`rag_ingest`	A document file on disk should be learned
`rag_ingest_base64`	A Slack attachment should be learned (decode → ingest)
`rag_ingest_url`	The user asks to ingest a web page (ALWAYS preview scope first)
`rag_search`	Question concerns vendor procedures, customer standards, install steps, or ingested content
`rag_list`	User asks what the knowledge base contains
`rag_stats`	User asks about corpus size/health or retrieval telemetry
`rag_update_metadata`	Fix a document's doc_type/title/version
`rag_delete`	User asks to remove a document (CONFIRM with the user first)
`rag_reindex`	Chunking/embedding config changed (CONFIRM with the user first)
`rag_snapshot`	User EXPLICITLY asks to store live output for later comparison (confirm scope first — never automatic)

The Four Knowledge Sources (routing rules)

Route every question to the right source. Most questions need NO retrieval.

Parametric knowledge — timeless networking fundamentals (OSPF LSA types, BGP path selection). Answer directly. Do not search.
Memory MCP (memory_recall, memory_get_facts, memory_get_decisions) — NetClaw's own past sessions, learned facts, and decisions about THIS network. "What was that BGP issue last month?" goes here, never to rag_search.
RAG knowledge base (rag_search) — user-uploaded documents. Vendor procedures, customer standards, install steps. Check it BEFORE declaring ignorance on these topics.
Live MCP servers (pyATS, NetBox, etc.) — current network state. NEVER answer a live-state question from the RAG store. The only exception is an explicitly requested snapshot, whose age must always be shown.

When a question legitimately touches both Memory and the knowledge base, consult both — but attribute every part of the answer to its actual source. Memory recall is never presented as a document citation, and vice versa.

"Remember this document" → ingestion (RAG). "Remember that PE2 is in maintenance until Friday" → Memory (memory_record_fact).

Workflow: Slack Document Upload

When a user posts a file attachment in the Slack channel and asks NetClaw to learn it:

Download the attachment content.

Call the ingest tool with the base64-encoded file:

bash

python3 $MCP_CALL "python3 -u $RAG_MCP_SCRIPT" rag_ingest_base64 \
  '{"filename": "wlc-9800-upgrade-guide.pdf", "content_base64": "<b64>", "doc_type": "vendor"}'

Infer doc_type from the user's words ("this is our customer standard" → customer); default other.
Confirm in-thread with what was learned: title, doc_type, page count, chunk count, collection — plus the example_question from the response, so the user learns what they can now ask.
If the response is deduplicated: true, say the document was already indexed. If reindexed: true, say the prior version was replaced.
Errors (unsupported format, size cap, parse failure) are reported verbatim — never silently swallowed.

Supported formats: PDF, Markdown, HTML, TXT, DOCX, XLSX, PPTX, VSDX natively; legacy DOC/XLS/PPT/VSD when LibreOffice is installed.

Workflow: URL Ingestion (always preview first)

When a user asks to ingest a web page:

Preview: rag_ingest_url {"url": "...", "mode": "preview"} — returns the page title, the same-domain pages it links to (depth 1, capped at RAG_CRAWL_MAX_PAGES), and a scope_token. No ingestion happens.
Confirm scope with the user: "That page links to 6 same-domain pages. Ingest just the page, or all 7?" ALWAYS offer the single-page fallback. If the preview was truncated, say so.
Ingest:
- Single page: rag_ingest_url {"url": "...", "mode": "ingest"}
- Page + linked pages (only after explicit confirmation): rag_ingest_url {"url": "...", "mode": "ingest", "include_linked": true, "scope_token": "<from preview>"}
Confirm as with file ingestion. Each page records its own URL as source. PDFs served at URLs are handled automatically.

Never call mode="ingest" with include_linked=true without having shown the preview and obtained the user's confirmation — the server rejects a missing/stale scope_token.

The Agentic Retrieval Protocol

Retrieval is a tool you wield, not a pipeline you sit inside. For every question:

1. Route first (Adaptive RAG)

Decide whether to retrieve AT ALL using the four-source rules above. Most questions need no retrieval:

Question	Route	Why
"What's the OSPF LSA type for external routes?"	Answer directly	Timeless fundamental
"What was that BGP issue last month on PE2?"	`memory_recall`	Your own past experience
"What does our customer standard require for change windows?"	`rag_search` (filter `doc_type: customer`)	User-uploaded document
"What's the current BGP state on PE2?"	Live MCP (pyATS)	Live network state — NEVER RAG
"Upgrade the lab WLC per our standards"	`rag_search` for procedure/standards, THEN live MCP for pre-checks	Mixed — each part to its source

2. Rewrite and decompose

Rewrite conversational phrasing into retrieval-friendly queries ("how do I get the new code on the WLC" → "WLC software upgrade procedure install activate commit"). Decompose multi-part questions into independent sub-queries, retrieved separately with scoped filters, then synthesized. Give each sub-query a sub_query_id and pass round on every rag_search call so the budget is auditable.

Show full SKILL.md (531 more words)Show less

3. Critique after every retrieval (Self-RAG)

Grade the returned chunks: do they actually answer the question?

Yes → answer, citing every claim.
Partially / no → rewrite the query (different terms, add filters) and re-retrieve.
low_confidence: true on results → treat as signal, never as answer material.

4. Respect the iteration budget — 3 rounds per sub-query

Each sub-query gets at most 3 retrieval rounds (initial + 2 refinements; RAG_MAX_ROUNDS). Stop condition: when a sub-query's budget is exhausted, stop retrieving for it and state plainly what you found (cited) and what remains unanswered. Do not loop. Do not pad.

5. Honest miss

If the corpus doesn't cover the topic — empty results, corpus_empty: true, or only low-confidence chunks after refinement — say so:

"The knowledge base doesn't cover Nexus 9300 upgrades. I can ingest a document if you have one — drop it in this channel."

NEVER answer from irrelevant or low-confidence chunks. NEVER present a guess as knowledge-base content. A fabricated answer is worse than an honest gap.

Citation Rules (non-negotiable)

Every claim derived from retrieved content carries a citation:

[WLC 9800 Upgrade Guide §4.2, p.31 — ingested 2026-07-01]

Use the citation field returned by rag_search verbatim. A claim you cannot attribute to a specific retrieved chunk is NOT presented as coming from the knowledge base. Chunk IDs stay in the retrieval log — never show them to users.

When synthesizing across multiple documents, every combined claim must be supportable by at least one cited chunk. Do not blend unrelated chunks into a claim none of them supports — that is synthesis hallucination.

Snapshots — the ONLY sanctioned RAG use of live data

rag_snapshot exists for one purpose: explicit point-in-time comparison ("snapshot the BGP tables so we can compare next month"). It is NEVER a substitute for a live MCP query.

Absolute prohibition: NEVER invoke rag_snapshot automatically, from a heartbeat, on a schedule, or as a side effect of another task. It runs only on an explicit human request, and only after you confirm scope.

Workflow:

User explicitly asks to snapshot live output.
Confirm scope before executing (HIIL gate): which devices, which commands, what label. "Scope: PE1, PE2 — show ip bgp via pyATS, into snapshot_core-bgp_<timestamp>. Confirm?"
Collect the output via the existing MCP tools (pyATS etc.).
Call rag_snapshot {"label": "core-bgp", "content": "<output>", "source_description": "core router BGP tables", "devices": ["PE1","PE2"], "commands": ["show ip bgp"]}.
Report what was stored including the per-type redaction counts from the response — secrets (passwords, SNMP communities, auth keys, pre-shared keys) are scrubbed before vectorization, and a zero count is reported explicitly.

When ANY answer later uses snapshot data:

Lead with the staleness notice: "From snapshot core-bgp — captured 2026-07-16 14:02 UTC — 31 days ago (live state is available via MCP tools): …" Use the age_human/staleness_notice fields verbatim.
Remind the user that live state is one MCP call away.
Snapshots older than RAG_SNAPSHOT_WARN_DAYS (default 90) are flagged stale — surface the flag. They are never auto-deleted; offer deletion instead.

Environment Variables

Variable	Default	Purpose
`RAG_DATA_DIR`	`~/.openclaw/rag`	Persistent store (never `~/.openclaw/memory/`)
`RAG_EMBEDDING_MODEL`	`BAAI/bge-small-en-v1.5`	Local embedding model
`RAG_RERANKER_MODEL`	`cross-encoder/ms-marco-MiniLM-L-6-v2`	Local reranker
`RAG_RERANK_ENABLED`	`true`	Disable on low-resource hosts
`RAG_MAX_DOC_MB` / `RAG_MAX_DOC_PAGES`	`100` / `1000`	Per-document ingestion caps
`RAG_CRAWL_MAX_PAGES`	`30`	Depth-1 crawl preview bound
`RAG_SNAPSHOT_WARN_DAYS`	`90`	Snapshot staleness warning
`RAG_MAX_ROUNDS`	`3`	Retrieval rounds per sub-query
`RAG_MCP_SCRIPT`	—	Path to `rag-mcp/rag_mcp_server.py` for `$MCP_CALL`

Example Usage

text

User: (attaches customer-wlan-standard.docx) learn this, it's our customer standard
NetClaw: Learned "Customer WLAN Standard" (customer, 24 pages, 61 chunks, documents).
         Try asking: "What does our standard 'Customer WLAN Standard' say about maintenance windows?"

Files

Just SKILL.md in workspace/skills/rag of automateyournetwork/netclaw.

Open the folder on GitHubat commit 95bb17e

Compare with similar skills

RAG next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

RAG compared with similar skills
Skill	Stars	Used in	Tokens	Auto-check	Licence	Repo updated
RAG this skillautomateyournetwork/netclaw	676	—	~2.8k	Automated safety check: Pass	Apache-2.0	3 days ago
Blockify Integrationiternal-technologies-partners/blockify-agentic-data-optimization	315	—	~6.2k	Automated safety check: Notes	Custom licence	5 mo ago
Agentsop Difyagentsope/SkillAlchemy	466	—	~5.4k	Automated safety check: Notes	MIT	today
Penguin SDKPrism-Shadow/penguin-harness	2.5k	—	~11k	Automated safety check: Pass	Apache-2.0	yesterday
Sc QAopen-edge-platform/edge-ai-suites	140	—	~2.3k	Automated safety check: Pass	Apache-2.0	today
RAG AssistantAtmosphere/atmosphere	3.8k	—	~504	Automated safety check: Pass	Apache-2.0	today

Similar skills

Blockify Integration
iternal-technologies-partners/blockify-agentic-data-optimization
Process documents with Blockify API to create optimized IdeaBlocks for RAG.
315 GitHub stars~6.2k tokensUpdated 5 mo ago
AI & LLM EngineeringAuto-check: notes
Agentsop Dify
agentsope/SkillAlchemy
SOP for building LLM applications on Dify — visual workflow + chatflow + agent + RAG knowledge base + plugin marketplace + observability, self-hostable.
466 GitHub stars~5.4k tokensUpdated today
AI & LLM EngineeringAuto-check: notes
Penguin SDK
Prism-Shadow/penguin-harness
A skill your agent uses whenever the user wants to build an agent application — their own program with an embedded agent, such as an AI app, an agentic app or a RAG app.
2.5k GitHub stars~11k tokensUpdated yesterday
AI & LLM EngineeringAuto-check passed
Sc QA
open-edge-platform/edge-ai-suites
Ask a natural-language question against indexed content via the Content Search RAG Q&A endpoint.
140 GitHub stars~2.3k tokensUpdated today
AI & LLM EngineeringAuto-check passed
RAG Assistant
Atmosphere/atmosphere
Knowledge base assistant that retrieves and cites documents from a curated index.
3.8k GitHub stars~504 tokensUpdated today
AI & LLM EngineeringAuto-check passed
Langchain4j RAG Implementation Patterns
giuseppe-trisciuoglio/developer-kit
Provides Retrieval-Augmented Generation (RAG) implementation patterns with LangChain4j for Java.
356 GitHub stars~3.3k tokensUpdated 29 days ago
AI & LLM EngineeringAuto-check: notes

Questions about RAG

What does RAG do?

Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base. RAG is an agent skill from automateyournetwork/netclaw. Ingest user-selected documents and retrieve cited procedures, standards, and design evidence from the local RAG knowledge base.

When should I use RAG?

RAG fits situations like: document questions; corpus management; explicitly requested snapshots.

How do I install RAG in Claude Code?

Run `npx skills add automateyournetwork/netclaw --skill rag -a claude-code`. Or copy the skill folder (workspace/skills/rag in automateyournetwork/netclaw) into .claude/skills/rag in your project. Claude Code loads it when a task matches its description.

How do I install RAG in Codex?

Run `npx skills add automateyournetwork/netclaw --skill rag -a codex`. Or copy the skill folder (workspace/skills/rag in automateyournetwork/netclaw) into .agents/skills/rag in your project. Codex loads it when a task matches its description.

Can I use RAG in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add automateyournetwork/netclaw --skill rag -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/rag, .gemini/skills/rag, .github/skills/rag and .opencode/skills/rag in your project.

What does RAG need to run?

Going by SKILL.md and its folder, RAG needs the command-line tools its instructions call (python3). Our summary lists: Python 3.

Does RAG access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is RAG safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does RAG use?

RAG is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does RAG use?

About 2.8k tokens (SKILL.md is roughly 11k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to RAG?

Skills that share tags, products or a category with RAG: Blockify Integration (iternal-technologies-partners/blockify-agentic-data-optimization, 315 stars), Agentsop Dify (agentsope/SkillAlchemy, 466 stars), Penguin SDK (Prism-Shadow/penguin-harness, 2.5k stars) and Sc QA (open-edge-platform/edge-ai-suites, 140 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains RAG?

automateyournetwork (a GitHub user) maintains it in automateyournetwork/netclaw, which has 676 GitHub stars. The repository holds 120 skills in this directory. The repository was last updated on October 5, 2026.

Source: automateyournetwork/netclaw on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.

Install RAG

Facts

At a glance

What it does

When your agent uses it

Example prompts

Requirements

Workflow steps

What it can do on your machine

Tool permissions

Runs code

Network

Credentials

Context cost

Safety

SKILL.md

Skill: RAG Knowledge Base

Overview

MCP Tools

The Four Knowledge Sources (routing rules)

Workflow: Slack Document Upload

Workflow: URL Ingestion (always preview first)

The Agentic Retrieval Protocol

1. Route first (Adaptive RAG)

2. Rewrite and decompose

3. Critique after every retrieval (Self-RAG)

4. Respect the iteration budget — 3 rounds per sub-query

5. Honest miss

Citation Rules (non-negotiable)

Snapshots — the ONLY sanctioned RAG use of live data

Environment Variables

Example Usage

Files

Compare with similar skills

Similar skills

More from automateyournetwork/netclaw

Related

Questions about RAG