Weaviate Database Operations Skill
Lets your AI agent search, query, and manage Weaviate vector database collections — semantic, hybrid, and keyword search plus AI-answered questions and data import.
Evidence shows clear disclosure of data flows for env vars, API keys, and external provider headers (references/environment_requirements.md lists which headers go to Weaviate); license and attribution are clear (Weaviate B.V., BSD-3-Clause); no covert exfiltration signs. Deducted for: no user-confirmation mechanism or rollback notes for write operations (collection creation, data import); no least-privilege/isolation guidance for API keys handled by scripts; actual script source not available for static audit. Not at the 25-point completeness bar.
Docs are internally consistent: parameter tables, usage, and examples align across references; a dedicated error-handling section maps common errors to remediation steps. Deducted for: the actual Python scripts (scripts/*.py) are absent from the evidence, so script self-consistency and behavior on abnormal input cannot be verified; no test evidence; dependency availability (uv, Python 3.11+) unverified. Static cap of 10 applies.
Trigger conditions are precise: each reference has explicit 'Use when' guidance, hybrid search is declared the default, search-type selection guidance is clear, and output formats are declared. Deducted for: core function depends entirely on Weaviate Cloud (an overseas service) with no mainland-China reachability note; no Chinese-language support; non-fit boundaries only partially declared (e.g., whether self-hosted Weaviate is supported is unstated).
Information architecture is well layered: SKILL.md index with progressive disclosure into references; install notes, env vars, output formats, and error handling present; README includes quickstart and a security note; BSD-3-Clause license explicit. Deducted for: no version number or changelog, no known-limitations section, maintenance responsibility and update path only implicit, hidden assumptions (uv required) not explicitly disclosed.
The skill claims end-to-end coverage of search, query, collection management, and data import; the recommended workflow (list before operating) is sensible and outputs are directly usable formats (Markdown tables/JSON). Deducted for: no execution evidence in static review, so core value (Query Agent answer quality, PDF import correctness) is unverified; depends on a paid cloud service and external embedding providers with no cost/benefit disclosure. Static cap of 7 applies.
Evidence consists of auditable primary documents and license text with clear fact/inference separation, plus official documentation links for cross-reference. Deducted for: no committed test suite or CI execution evidence covering the skill's key paths, no third-party verification, limited cross-source corroboration. Static review without test evidence caps this at 5.
- The skill depends entirely on Weaviate Cloud (an overseas service); mainland-China users should assess network reachability before use.
- All scripts rely on Python source not shown in the evidence; review each script's actual behavior in scripts/ before use.
- Collection creation and data import are write operations without confirmation or rollback mechanisms; validate in a test environment first.
- API keys (WEAVIATE_API_KEY and provider keys) pass through the scripts; mind key-exposure risk and set only required keys.
- No version numbers or changelog, making it hard to track skill changes.
- This is a static documentation review only; no scripts were executed and effectiveness is independently unverified.
What does this skill do, and when should you use it?
This is the Weaviate-focused skill from Weaviate's official agent-skills repository, giving agents database access through a set of Python scripts. It supports multiple search modes (semantic, hybrid, keyword, and the Query Agent's ask and search modes), collection management (listing, inspection, creation, data exploration), and imports from CSV, JSON, JSONL, and PDF files. The skill connects to a Weaviate Cloud instance configured via environment variables, and external vectorizer provider keys are auto-detected. It works with any agent client that supports the Agent Skills standard, such as Claude Code, Cursor, and Gemini CLI.
Runs Python scripts in the scripts/ directory (via uv) to: perform semantic, hybrid, and keyword search across collections; use the Query Agent in ask mode (structured answers with source citations — collection name and object ID) or search mode (raw objects); list collections and inspect their schema (properties, vectorizer configuration, replication factor, multi-tenancy status); explore data distribution and statistics; fetch objects by ID or strict filters; create collections with custom schemas; import data from PDF/CSV/JSON/JSONL (PDF imports auto-create the collection); and generate example data. Output supports Markdown tables (default) and JSON (-- flag).
- Developers wiring AI applications to Weaviate who want agents to handle search and data operations from natural language instructions instead of hand-written SDK code
- Data engineers who need to quickly load local CSV, JSON, JSONL, or PDF documents into a Weaviate collection and start retrieving
- Teams building RAG pipelines who want the Query Agent to return direct answers with source citations (collection and object ID)
- New users without a Weaviate instance who want to get started in a free sandbox, using the quickstart onboarding and example data generation
- Ops or data analysts who need to inspect collection schemas, data distribution, and statistics to validate data quality before querying
What are this skill's strengths and limitations?
- Officially maintained by Weaviate with deep integration into the Query Agent and Weaviate Cloud
- Covers the full workflow end to end: collection creation, data import, multi-mode search, Q&A, and data exploration
- Clear guidance for choosing the right search strategy reduces misuse
- Uniform output in Markdown tables or JSON, suitable for both agent display and programmatic use
- Built on the open Agent Skills standard, portable across Claude Code, Cursor, Gemini CLI, and other clients
- Hard dependency on a Weaviate Cloud instance and network connectivity; the documentation does not describe a path for offline or self-hosted Weaviate
- External vectorizer provider API keys may incur additional costs that are not documented
- The repo has no topics set and no documented test suite, leaving limited quality-assurance evidence
- Requires comfort with environment variable setup; the full list of provider keys lives in a separate reference document
- Collection creation defaults to the text2vec_weaviate vectorizer; custom embedding models require reading the environment requirements doc
How do you install this skill?
Option 1 (recommended; works with Cursor, Claude Code, Gemini CLI, etc.): run npx skills add weaviate/agent-skills. Option 2 (Claude Code plugin manager): run /plugin marketplace add weaviate/agent-skills then /plugin install weaviate@weaviate-plugins. Option 3 (manual): git clone https://github.com/weaviate/agent-skills.git and point your agent at the directory. Note the repo bundles 2 skills; this one lives at skills/weaviate/. After installing, you need Python 3.11+, uv, a Weaviate Cloud instance, and the WEAVIATE_URL and WEAVIATE_API_KEY environment variables; external vectorizer provider keys are auto-detected.
How do you use this skill?
The skill is auto-discovered by compatible agents — just describe what you want, e.g. "Search my documentation collection for information about HNSW indexing" or "Import data.csv into the Products collection". If unsure what collections exist, the agent starts with uv run scripts/list_collections.py. Claude Code users can also use slash commands: /weaviate:quickstart for interactive onboarding, or /weaviate:search query "best laptops" collection "Products" type "hybrid" for hybrid search. Search type guidance: general search → hybrid (default), conceptual similarity → semantic, exact terms/IDs → keyword, cross-collection answers → ask, cross-collection raw browsing → query_search.
How does this skill compare with similar options?
The same repository also bundles a Cookbooks skill offering end-to-end application blueprints (multimodal PDF ingestion, basic/advanced/agentic RAG, Query Agent chatbot, frontend interface) — better if you want to build complete apps; this skill focuses on direct database operations. If you don't use Weaviate, this skill won't help — look for a skill for your database or use Weaviate's official SDK directly.