Claude Skill MartBrowse skillsQuick linesLearn by videoTerminal guideWhat is a Skill?
 
← Back to list

Human in the Loop (Approval Gate Design)

Designs and verifies auditable human approval gates, escalation paths, and safe state transitions for risky AI agent actions.

Security & ReviewAdvanced★ 207⑂ 42AI score 9/10Last updated: Aug 9, 2026

What it does

  • Inventories every action an agent can take and assigns a risk tier based on reversibility, value, blast radius, and data sensitivity.
  • Picks the lightest sufficient control: autonomous-with-audit, notify-after, review-before, step-up approval, dual control, or prohibited.
  • Specifies an explicit state machine (prepared -> pending_review -> approved | rejected | expired) plus expiry, timeout-to-deny, escalation, compensation, and break-glass paths.
  • Designs the approver's decision package (exact action, before/after diff, provenance, uncertainty, safe reject/edit) and an append-only decision record schema with retention/redaction rules.
  • Ships assets/approval-policy-template.json, a validator at scripts/validate_gate_policy.py, and design guidance in references/gate-design-guide.md.

Who it's for

  • Engineers and architects building agents with real execution power: payments, outbound email, deploys, access grants.
  • Platform, security, and compliance teams that need separation of duties and audit trails.
  • Teams upgrading a bare "confirm?" prompt into a defensible approval system.

Example uses

  1. Refund agent: auto-refund under $50 for verified duplicate charges, support manager above $50, finance plus support above $1,000, with approval bound to order, amount, destination, and policy version.
  2. Customer comms agent: agent drafts, account owner approves exact recipients, subject, body digest, attachments, and send time; any edit invalidates the approval.
  3. Audit of an existing workflow: generate tests proving self-approval, duplicate callbacks, and expired approvals cannot execute twice.

· · · Install guide · · ·

Try it now, no install

Paste this into Claude to use the skill without installing anything.

Read the instructions in this file and follow them to help me:
https://raw.githubusercontent.com/seb1n/awesome-ai-agent-skills/HEAD/agent-engineering/human-in-the-loop/SKILL.md

What I want: (describe your task here)

If Claude can't open the link, open it yourself and paste the contents instead.

↓ If it works for you, download the ZIP below and install it. Then it runs on its own — no pasting each time.

Install in the Claude app (no terminal)
  1. Download the ZIP with the button below.
  2. In Claude, open Settings → Capabilities and turn on 'Code execution and file creation'. (one time)
  3. Go to Customize → Skills → + → 'Upload a skill' and upload the ZIP.
↓ Download ZIP
Install in Claude Code

Let Claude do it — paste this into Claude Code

Install the skill I found on Claude Skill Mart.
Copy the agent-engineering/human-in-the-loop folder from the GitHub repo seb1n/awesome-ai-agent-skills into my ~/.claude/skills/human-in-the-loop/.
When it's done, tell me in one line what this skill can do.

Install with a command instead

git clone https://github.com/seb1n/awesome-ai-agent-skills.git /tmp/aas && mkdir -p ~/.claude/skills && cp -r /tmp/aas/agent-engineering/human-in-the-loop ~/.claude/skills/

⚠ This is a third-party skill. Check the source repository before installing.

  1. Open a terminal.
  2. Create the skills folder: mkdir -p ~/.claude/skills
  3. Clone the repository: git clone https://github.com/seb1n/awesome-ai-agent-skills.git /tmp/aas
  4. Copy just this skill: cp -r /tmp/aas/agent-engineering/human-in-the-loop ~/.claude/skills/
  5. Verify the files: ls ~/.claude/skills/human-in-the-loop (expect SKILL.md, assets, scripts, references).
  6. Restart Claude Code, then ask something like "design an approval gate for my agent's payment action".
  7. Install Python 3 if you want to run the policy validator: python3 --version
View source on GitHub ↗License: MIT