Moved: this skill now lives in cbzehner/skills under skills/agent-ergonomic-design/. This repo is archived and read-only.

Agent Ergonomic Design

Design CLIs, APIs, and MCP tools that agents can call without guessing. The skill focuses on machine-readable output, clear failure modes, and interfaces that work well in real agent loops.

Use It For

• Adding robot or machine mode to a CLI
• Reviewing an API or MCP tool for agent use
• Turning human-only output into structured output an agent can trust

Install

Clone the repo and run the installer:

git clone https://github.com/cbzehner/skill-agent-ergonomic-design.git
cd skill-agent-ergonomic-design
./install.sh all

Install targets:

• ./install.sh claude installs to ~/.claude/skills/agent-ergonomic-design
• ./install.sh codex installs to ~/.codex/skills/agent-ergonomic-design
• ./install.sh agents installs to ~/.agents/skills/agent-ergonomic-design
• ./install.sh opencode installs to ~/.config/opencode/skills/agent-ergonomic-design
• ./install.sh all --copy copies files instead of symlinking

Manual install works too: symlink or copy skills/agent-ergonomic-design into your agent's skills directory.

Agent Support

This repo uses the plain skills/agent-ergonomic-design/SKILL.md layout. Claude Code and Codex also get small plugin manifests at .claude-plugin/plugin.json and .codex-plugin/plugin.json.

Other agents can read the same SKILL.md file. If a host does not support a frontmatter field or tool name, ignore that field and follow the workflow text.

Layout

.claude-plugin/plugin.json
.codex-plugin/plugin.json
install.sh
skills/agent-ergonomic-design/SKILL.md
README.md
LICENSE

Public Notes

These repos are public. Keep private repo names, secrets, customer data, raw logs, cookies, and absolute filesystem paths out of examples.

License

MIT