Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+22 moreClaude CodeCursorCodex CLIMuseDots
OpenClaw
Manus
Grok
Grok Bot+18 agentsHallucination Guard — Design AI Systems
Design controls that reduce unsupported AI claims through evidence checks, uncertainty labels, action verification and evaluation gates.
$9.99
Hallucination Guard — Design AI Systems
Example session with this skill installed
Fictional example: A customer-service bot invents order statuses when its order API is unavailable. We need a safer response policy. Produce a complete first-pass deliverable, separate facts from assumptions and give practical next steps. Do not take external actions.
- Read your context and instructions
- Compiled the hallucination guard
Hallucination Guard: worked example
Scope: this fictional brief, using only supplied information.
A customer-service bot invents order statuses when its order API is unavailable. We need a safer response policy.
Deliverable
| Workstream | Finding / decision | Recommended action |
|---|---|---|
| Authority | Order API is the status source | Require a successful authorized lookup before stating order status; model text cannot substitute for missing account evidence. |
| Failure response | Unavailable API means unknown status | Use: I cannot verify the current order status right now. Please retry or contact support with your order reference. Do not claim shipped or delivered. |
| Evaluation | Need both quality and failure coverage | Test unavailable API, stale cache, wrong account, partial response and conflicting status; assert no invented order facts or completed actions. |
Decision and next steps
Implement the evidence gate and observe abstention quality and resolution rate. Human escalation must preserve context without exposing unnecessary data. No order lookup or code deployment occurred.
Evidence boundaries
Conclusions apply to the supplied scenario. Missing facts remain unverified; recommendations are not completed actions. Validate source data and approvals before implementation. Track results against the objective and adjust only when new evidence supports change.
Connects securely to your tools. The creator never sees your data.
About this skill
Hallucination Guard
Design controls that reduce unsupported AI claims through evidence checks, uncertainty labels, action verification and evaluation gates.
Workflow and deliverables
- Require a successful authorized lookup before stating order status; model text cannot substitute for missing account evidence.
- Use: I cannot verify the current order status right now. Please retry or contact support with your order reference. Do not claim shipped or delivered.
- Test unavailable API, stale cache, wrong account, partial response and conflicting status; assert no invented order facts or completed actions.
Getting started
Provide your goal, relevant source material, constraints and context for hallucination guard. Include dates, existing decisions and any restrictions that must be preserved.
Receive a structured deliverable with source facts, explicit assumptions, priorities, open questions and next steps. Adapt the depth to your available evidence; the sample below demonstrates the workflow with fictional inputs.
Boundaries
Uses supplied evidence and explicit assumptions. No guaranteed outcomes or continuous monitoring. Live research and external actions require supported tools and authorization.
Reusable agent instructions, with optional tool use depending on your host. External actions require supported tools and explicit authorization.
How to use
Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill passes our trust checks below.
Security scanned
Passed 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 scanned
Verified clean 6 days ago
- Passed all security checks, Safe to install