More screenshots
Works with the AI Agents you already use
TenantGround (Document Chatbot Builder)
Build a company-document chatbot with source-backed answers, clear no-answer behavior, and optional multi-tenancy. Includes synthetic documents, acceptance cases, and a local validation workflow.
$4.99
TenantGround (Document Chatbot Builder)
Example session with this skill installed
Use TenantGround's bundled offline example to demonstrate a private document chatbot for Company A and Company B.
Show each company's return policy with its supporting source, refuse a question that the documents do not answer, and verify restricted HR access and conversation boundaries. Run the multi-company and single-company acceptance checks and the application tests.
Use only the supplied synthetic documents. Keep everything local, with no external model calls or deployment. Report actual results and limitations.
- Read your context and instructions
- Compiled the tenantground document chatbot
TenantGround: recorded offline demonstration
This report uses actual execution records from the bundled offline reference on 28 October 2026. It is a demonstration of the included example, not a live LLM conversation or a cloud deployment. Local machine paths have been omitted.
Delivered example
The reference imports fictional Markdown documents into SQLite, scopes evidence to the authenticated company/bot/group, returns conservative extractive answers with source excerpts, and includes a small browser chat UI. It supports both one-company and multi-company configurations.
Observed checks
| Check | Recorded result |
|---|---|
| Multi-company acceptance | 17/17 passed |
| Single-company acceptance | 12/12 passed; 5 not applicable |
| Application tests | 18/18 passed |
| Extracted-package verification | All six runner steps exited successfully |
| Corpus and retrieval | 8 fictional documents, 20 chunks, top-k 3 |
Actual answer and source examples
Company A: return policy
Outcome: answered
Answer: The return window is 7 days after receipt. [1]
Source: A-Handbook.md — Return policy (a-returns)
The return window is 7 days after receipt.
Company B: return policy
Outcome: answered
Answer: The return window is 30 days after receipt. [1]
Source: B-Handbook.md — Return policy (b-returns)
The return window is 30 days after receipt.
Unsupported discount question
Outcome: refused
Answer: The authorized documents do not contain enough information to answer this question.
Sources: none.
Private cross-company request
Outcome: denied
Answer: Access denied.
Sources: none.
Complete multi-company case outcomes
| Case | Observed outcome | Contract |
|---|---|---|
| a-return | answered | Passed |
| b-return | answered | Passed |
| a-leave | answered | Passed |
| b-leave | answered | Passed |
| unsupported-discount | refused | Passed |
| unsupported-bonus | refused | Passed |
| cross-tenant-access | denied | Passed |
| spoofed-tenant | denied | Passed |
| wrong-bot-conversation | denied | Passed |
| wrong-owner-conversation | denied | Passed |
| restricted-hr | refused | Passed |
| authorized-hr | answered | Passed |
| wrong-bot-document | refused | Passed |
| right-bot-document | answered | Passed |
| source-trimming | answered | Passed |
| current-ready-only | answered | Passed |
| injection-does-not-set-policy | answered | Passed |
The 18 application tests additionally exercise real text ingestion, failure/retry, immutable versions, replacement rollback, session expiry, concurrent company-scoped requests, source trimming, and local HTTP login and identity-spoofing behavior.
Reproduce
Extract the ZIP and run from the folder containing SKILL.md:
python examples/verify.py --skill .
python examples/app.py --config examples/configs/multi.json serve
Open http://127.0.0.1:8765. The example guide lists the synthetic demo accounts.
Next steps and limitations
Adapt the workflow to your own application, authentication, document parser, retrieval store, and model provider. Then test those integrations with your own authorized data.
This reference uses lexical search and quotes approved source text. It does not test LLM generation, embeddings, PDF/OCR, multilingual answer behavior, cloud deployment, or production security. The fixture checker verifies source/access contracts and selected expected values; it is not a semantic faithfulness judge or a guarantee against hallucinations.
Connects securely to your tools. The creator never sees your data.
About this skill
Build a document chatbot with your coding agent, using your existing or chosen stack.
TenantGround guides the complete path from business documents to authorized retrieval, source-backed answers, and clear refusals when information is missing. Use it for one company's internal assistant or an application serving multiple clients.
What you receive
- A focused English SKILL.md with a provider-neutral implementation workflow.
- An implementation contract for document, bot, conversation, and access boundaries.
- Fictional documents for two companies and 17 acceptance scenarios.
- A Python checker for exported response captures.
- An optional runnable offline example with a chat UI and application tests.
- Reproduction commands and validation notes explaining what was actually tested.
The recorded offline example passed 17 multi-company cases, 12 single-company cases, and 18 application tests. Five second-company cases are explicitly not applicable in single-company mode. These are local fixture results, not a live-model accuracy score.
Who it is for
Developers, freelancers, and agencies building internal document assistants or customer-support chatbots.
Where it stops
This is a coding-agent workflow with a reference example. Hosting, production authentication, database/model services, OCR, and recurring billing are not supplied. The offline example uses lexical retrieval and extractive answers; it does not call an LLM or generate embeddings. Test your selected production stack separately.
About the screenshots
Screenshots show HelpBot, the creator's existing RAG application used as a reference case study for TenantGround. HelpBot's dashboard, PDF pipeline, and styling are not included in the ZIP. TenantGround includes a separate offline example; your application's features and design depend on the implementation you request.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
The bundled reference imports local Markdown/text and uses lexical retrieval with extractive answers; it does not call an LLM or generate embeddings. PDF parsing, OCR, production authentication, cloud hosting, recurring billing, and team administration require separate implementation. Local synthetic test results do not establish live-model accuracy, production security, or compatibility with every agent/provider. Source references do not guarantee that every generated claim is correct.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 20 days ago
- Software
Python 3.10+
Python runs the bundled checker and the optional offline example.
Permissions
File Scopes
Reads and writes only within the project workspace selected by the user and uses the terminal for implementation and local checks. The complete offline example uses a temporary localhost HTTP listener. No connected account, API key, external model call, or environment-variable access is required for the bundled demo/checker. Production integrations may require the user's selected services and credentials. Do not expose secrets in reports or browser code.