HumanLayer Skills - Setup Guide¶
Source: humanlayer/skills (5 skills · 2.7K combined installs)
Repo: github.com/humanlayer/skills
Category: Agent Infrastructure / Human-in-the-Loop
First Seen: August 13, 2026
Quality Tier: 🟡 Beta core (show-me at 2.0K installs, trending on the hot leaderboard)
HumanLayer is one of the leading human-in-the-loop (HITL) tooling companies, and this skill pack encodes their patterns for designing agent systems where humans approve critical actions. show-me teaches agents to render work for human review; design-control-loop and build-iterated-agentic-loop encode the architecture patterns behind approval gates. Directly relevant to governance-layer design for multi-agent deployments.
Installation¶
# Full cluster
npx skills add humanlayer/skills
# Hermes: install individual skills by identifier
hermes skills install humanlayer/skills/show-me
hermes skills install humanlayer/skills/design-control-loop
Core Skills¶
| Skill | Installs | Use For |
|---|---|---|
show-me |
2.0K | Rendering agent work for human review and approval |
improve-claude-md |
491 | Improving agent memory/context files |
design-control-loop |
92 | Designing human approval loops around agent actions |
build-iterated-agentic-loop |
74 | Building iterated agent execution with checkpoints |
narrow-react-prop-types |
45 | Tightening React props for HITL UI components |
Prerequisites¶
| Requirement | Details |
|---|---|
| Any agent runtime | Patterns are framework-agnostic |
| Node.js + npx | For the skills.sh CLI install path |
| An approval UI or channel | HumanLayer SDK, Slack, or custom review surface |
CorpusIQ Use Cases¶
| Use Case | How |
|---|---|
| Commander approval gates | design-control-loop patterns for the submit() review cycle |
| Customer deployments | HITL review surfaces for client Hermes installations |
| Audit trails | Human approval points as governance evidence |
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations / Verification¶
- Cluster is small and beta-tier overall -
show-meis the proven core - HITL design needs a review surface to be meaningful; the skill supplies the pattern, not the UI
- Verify install:
npx skills list | grep -E 'show-me|control-loop'