PlugWiki

Agent Platform RAG Engine Management Skill from Google: List, Inspect and Query RAG Corpora

Google's own skill for listing, inspecting and querying RAG Engine corpora on Google Cloud, and for answers grounded in a corpus.

Installs
9K
on skills.sh
Stars
21.8K
on GitHub
Updated
Nov 2026
last repo commit

agent-platform-rag-engine-management makes the agent use the vertexai Python SDK to list RAG corpora and files, inspect a corpus, retrieve matching contexts for a query, and generate an answer grounded in a corpus. Listing, inspecting and retrieving run without asking you.

Grounded generation asks you to confirm first, with the corpus ID, model and query shown. You need a Google Cloud project and gcloud sign-in. It is not for standard database queries, Google Workspace RAG or gRAG. It lives in Google's own skills repo and its folder is one file, SKILL.md.

Good for

  • Google Cloud developers who use Agent Platform (Vertex AI) RAG Engine
  • Engineers who need to find the right corpus or files before building on it
  • Teams testing grounded answers from a private corpus with a Gemini model

Install agent-platform-rag-engine-management

Claude Code

  1. 1

    Download the skill folder from https://github.com/google/skills/tree/HEAD/skills/cloud/agent-platform-rag-engine-management.

  2. 2

    Put it at ~/.claude/skills/agent-platform-rag-engine-management/ so that SKILL.md sits directly inside.

  3. 3

    Start a new session; the skill loads when its description matches the task.

No Google plugin in the repo's marketplace lists this skill, so install it as a skill. Before use you need gcloud auth login and gcloud auth application-default login, as the skill's Phase 0 says.

Built from the package data we hold. Not yet tested by us.

What's in the skill

When it loads

Manage and query Agent Platform RAG Engine Corpora and retrieve grounded contexts using the Google GenAI SDK. Use when listing RAG corpora or files, inspecting a corpus, retrieving contexts, or generating content grounded in a RAG corpus. Do not use for standard database queries (use SQL/Spanner skills), Google Workspace RAG, or other RAG products like gRAG.

Sections in SKILL.md

  1. Safety & Confirmation Tiers (CRITICAL)
  2. Phase 0: Environment Setup
  3. Workflow Decision Tree
  4. 1. Listing Corpora and Files (Discovery)
  5. 1.1 Listing and Discovering Corpora
  6. 1.2 Listing and Discovering Files
  7. 2. Getting / Inspecting an Agent Platform RAG Engine Corpus
  8. 3. Retrieving Contexts
  9. 4. Answering the User with Retrieved Context

Key facts

Author
Google LLC (owner named in the repo's marketplace.json; Google's own repo)
License
Apache-2.0 (repo LICENSE)
Installs on skills.sh
9,049 (Nov 2, 2026)
GitHub stars
21,764 for the google/skills repo (Nov 2, 2026)
Last commit
Nov 1, 2026
Files
1 file: SKILL.md; no scripts
Needs
gcloud sign-in with Application Default Credentials; Python packages google-cloud-aiplatform and google-genai

What it includes

  • Signs in with gcloud and application-default credentials; installs packages if missing

    Runs `gcloud auth login` and `gcloud auth application-default login`. If a snippet fails with ModuleNotFoundError, it runs pip install -q for google-cloud-aiplatform and google-genai and retries. No virtual environment is created.

    Runs code · Phase 0: set up

  • Discovers corpus resource names and files in a project and region

    Uses rag.list_corpora and rag.list_files. It discovers IDs itself before asking you. Tier R: no confirmation needed.

    Try: “List the RAG corpora in my project in us-central1”

    Network · 1. List corpora and files

  • Gets a corpus by name and prints its display name

    Uses rag.get_corpus. Tier R: no confirmation needed.

    Network · 2. Inspect a corpus

  • Runs a retrieval query and prints matching text and source URIs

    Uses rag.retrieval_query with similarity_top_k of 3 in the example. No results is a valid answer and is reported with the corpus and query.

    Try: “Search my corpus for the remote work policy”

    Network · 3. Retrieve contexts

  • Generates an answer with a Gemini model using the corpus as a tool

    Uses client.models.generate_content with a RAG store tool. Tier RC: you must say Yes after seeing the corpus ID, model and query, and the agent may not run it in the same turn as the prompt.

    Runs code · 4. Grounded answer

  • Tier R reads freely; Tier RC asks you first before using model compute

    The skill's own wording: Tier RC is read-only but consumes compute resources.

    Safety tiers

From the official docs

Permissions

Medium risk

The skill signs in to your Google Cloud account, reads your RAG corpora and files, and can run model generation after your confirmation. Use credentials limited to the project you mean to query.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • It runs gcloud logins for you

    Phase 0 uses `gcloud auth login` and `gcloud auth application-default login`, which set up credentials the SDK then uses. Check which account and project they sign in to. github.com

  • Retrieved text from your corpus enters the chat

    Retrieval prints context text and source URIs from your corpus, so private documents can appear in the conversation. github.com

  • Grounded generation asks first

    The skill requires a Yes with the corpus ID, model and query before generate_content runs. Read the parameters before approving. github.com

  • It may install Python packages

    Only when a snippet fails with ModuleNotFoundError, it runs pip install -q google-cloud-aiplatform google-genai. github.com

Technical details
Scopes

  • gcloud auth login and application-default credentials

  • Read RAG corpora, files and contexts

  • Model compute for grounded generation (asks first)

Questions

What does the RAG Engine management skill do?

It lets the agent list your RAG Engine corpora and files, inspect a corpus, retrieve contexts for a query and generate an answer grounded in a corpus. It uses Google's Python SDK.

Is it official?

Yes. It is in Google's own skills repository, and the marketplace file names Google LLC as owner.

How do I install it?

Use the install box above. The README installs with the skills CLI and asks which skills, agents and scope you want. No plugin in the repo's marketplace lists this skill.

What do I need before using it?

A Google Cloud project and region, a corpus, and gcloud sign-in with Application Default Credentials. The Python packages google-cloud-aiplatform and google-genai are installed only if a snippet fails without them.

Does it ask before generating answers?

Yes. SKILL.md requires an explicit Yes after the agent lists the corpus ID, model and query, and it must not run the generation in the same turn as the prompt. Listing and retrieval run without asking.

Who pays?

The skill has no price of its own. Generating content uses model compute in your Google Cloud project, which SKILL.md flags as consuming compute resources. Check Google Cloud's pricing for the model you use.

Is it safe?

It signs in to your Google Cloud account with gcloud and reads your RAG corpora. Generation needs your confirmation first. It may install two Python packages with pip, and SKILL.md notes a deprecation warning from vertexai.preview.rag is expected.

Information

updated Nov 2, 2026

Repository
google/skills
Folder
skills/cloud/agent-platform-rag-engine-management

google/skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2
  1. github.com · github.com
  2. github.com · github.com

Updated Nov 2, 2026.