mitidOfficielby @AgentAccess

MitID Human-in-the-Loop Safeguards

Deterministiske eksekveringsprotokoller til pausering af agentloops og håndtering af MitID-godkendelser med menneske-i-løkken.

Vis på GitHub

Installation

skills.sh / CLI

Installer denne færdighed direkte i din AI-agent eller projekt.

npx skills add lordavocado/AgentAccess.dk --skill mitid-hitl-safeguards

Om MitID Human-in-the-Loop Safeguards

Forhindrer autonome agenter i at gå i uendelige retry-loops, blokere brugerkonti eller forsøge at omgå nationale sikkerhedsporte (MitID, NemLog-in). Fastlægger strenge Human-in-the-Loop (HITL) afleveringsmønstre, tilstandsbevaring og timeouts.

SKILL.md Instruktioner

SKILL.md

De rå trinvise instruktioner som din agent udfører ved aktivering.

---
name: mitid-hitl-safeguards
description: Protocols for handling MitID authentication with mandatory human-in-the-loop safeguards.
metadata:
  compliance: mitid-hitl
  country: DK
---

# MitID Human-in-the-Loop (HITL) Safeguard Skill

Autonomous agents CANNOT solve MitID challenges, push notifications, or hardware authenticators.

## Safety Principles
1. **Zero Bypass Attempt**: Never simulate clicks on MitID authenticators or attempt credential stuffing.
2. **Immediate Pause**: The instant a MitID authentication challenge is generated, transition the agent state to `WAITING_FOR_HUMAN`.
3. **Clear Handoff Prompt**: Output a direct, unambiguous instruction to the human operator:
   > "MitID confirmation requested. Please open the MitID app on your mobile device and swipe to approve action: [action description]. Agent paused."
4. **Timeout Cap**: Poll for completion status at maximum every 5 seconds, up to 180 seconds. If the human does not approve within 3 minutes, gracefully terminate with `AUTH_TIMED_OUT`.

Relateret dansk system

MitID & MitID Erhverv

Den nationale danske digitale identitetsinfrastruktur for borgere og virksomheder.

Vis system