- Home
- Skills
- Productivity
- MindGuard Attention Firewall
Works with the AI Agents you already use
MindGuard Attention Firewall
A guardrail-driven attention decision skill that evaluates supplied usage signals, applies privacy and autonomy safeguards, and recommends explainable, reversible interventions while protecting essential apps.
$5
MindGuard Attention Firewall
Example session with this skill installed
Use the MindGuard Attention Firewall skill to run its bundled Python demonstration. Show the complete output for all three scenarios, explain the risk score, confidence, allowed actions, and selected action, and describe how the one-step escalation rule and essential-app protection affect the decisions. Clearly distinguish the built-in demonstration from real user activity. Do not invent usage data or claim that any application was actually monitored or blocked.
- Read your context and instructions
- Compiled the mindguard attention firewall
MindGuard Attention Firewall — Demo Results
Summary
The MindGuard reference engine evaluated three built-in decision scenarios. It produced an ALLOW decision for the low-risk scenario, a DELAY decision for the moderate-risk scenario, and an ALLOW decision for the essential-application scenario.
Complete Engine Output
Command
py -3 scripts\decision_engine.py --demo
Output
{"risk": 0.06, "confidence": 1.0, "allowed_actions": ["ALLOW", "NUDGE"], "action": "ALLOW", "reason": "Risk 0.06 (confidence 1.0); guardrail limit: escalate one rung at a time."}
{"risk": 0.588, "confidence": 1.0, "allowed_actions": ["NUDGE", "DELAY"], "action": "DELAY", "reason": "Risk 0.588 (confidence 1.0); guardrail limit: escalate one rung at a time."}
{"risk": 0.863, "confidence": 1.0, "allowed_actions": ["ALLOW"], "action": "ALLOW", "reason": "Risk 0.863 (confidence 1.0); guardrail limit: essential app is never restricted."}
Scenario Analysis
| Scenario | Risk | Confidence | Selected action |
|---|---|---|---|
| Low risk | 0.06 | 1.0 | ALLOW |
| Moderate risk | 0.588 | 1.0 | DELAY |
| Essential application | 0.863 | 1.0 | ALLOW |
1. Low-Risk Scenario
The risk score is 0.06, and the engine selects ALLOW. NUDGE is also available as an allowed action.
This demonstrates that low-risk activity does not automatically trigger restrictive intervention.
2. Moderate-Risk Scenario
The risk score is 0.588. The engine selects DELAY, with NUDGE and DELAY available as permitted actions.
The decision follows the gradual-intervention principle rather than immediately applying a stronger restriction.
3. Essential-Application Scenario
The risk score is 0.863, but the engine still selects ALLOW.
This is an important safety property: essential applications are protected even when the risk score is high.
What This Demonstrates
MindGuard does not rely on a risk score alone. Its guardrails constrain the available actions and ensure that essential applications remain accessible.
The system also provides explainable outputs containing the risk score, confidence, allowed actions, selected action, and reasoning.
Automated Test Results
The package includes eight unit tests covering:
- Invalid daily counts.
- Cooldown enforcement.
- Essential-app protection.
- Invalid input handling.
- One-level intervention escalation.
- Guardrail enforcement over preferred actions.
- Strict Boolean consent.
- Conservative handling of unknown history.
All eight tests passed against the final Agensi ZIP.
Ran 8 tests in 0.002s
OK
Next Steps
- Integrate the decision engine with a suitable source of user-provided attention signals.
- Define personalized attention rules and essential applications.
- Connect the recommended actions to an external application or device-control layer if enforcement is required.
- Test decisions with realistic scenarios before deployment.
Important: These are built-in demonstration scenarios, not results from live user monitoring. The package provides decision support and guardrails; external integrations are responsible for actual device enforcement.
Connects securely to your tools. The creator never sees your data.
About this skill
Purchase Model and Access
MindGuard Attention Firewall is offered as a one-time purchase, not a subscription. Buyers receive a permanent, personal licence to use the purchased skill through their Agensi-connected AI agent.
There are no recurring monthly or annual charges.
Future updates are not guaranteed or promised on a fixed schedule. If the creator publishes an update, Agensi automatically makes the updated version available to existing purchasers through their connected agent.
The purchase does not include permission to redistribute or resell the skill.
Requires host app to handle UI/OS-level blocking. Fails soft to ALLOW/NUDGE if state data is missing. Essential apps (e.g., banking, health) are hard-coded for exemption.
- This listing is a one-time purchase, not a subscription.
- Future updates are not guaranteed or included as an ongoing maintenance commitment.
- The purchase grants a personal, non-transferable licence through Agensi; it does not provide downloadable source files or redistribution rights.
- Actual device monitoring and operating-system enforcement require separate integrations.
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 any payment.
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 6 days ago
Requires
Permissions
File Scopes
Read access is for the bundled skill documentation, decision framework, guardrails, and examples. Terminal access is used to execute the included Python reference engine and tests. The package does not require external network access, account connections, environment variables, or writing to user files.