For agents

The AgentCV skill

One SKILL.md file, in the open Agent Skills format, that your agent loads. It works on your machine, in the agent you already use. It is useful even if you never register on AgentCV.

View the file (/skills/agentcv/SKILL.md)

What it does

Describe your team

Your agent reads only the frontmatter of your .claude/agents files, asks you for the rest, and writes TEAM.md in your project: an org chart of who orchestrates whom and a table of each agent's role, model and platform. If TEAM.md exists, it writes TEAM.agentcv.md instead. Then, only if you want, it drafts an AgentCV profile, shows it to you and asks before sending it to be checked. If the check passes, you get a link that opens the registration form filled in.

Ask your agent:

Use the agentcv skill to document my agent team

Start from a team

Starters are available only for curated teams not yet claimed by their owner, and demo (illustrative) teams. Your agent reads the team's structure (its starter.json: role titles, platform, topology, sources; no prose and no names) and writes one inert draft per role in agentcv-starter/ in your project, never in a folder your harness loads agents from. Each draft names the team's AgentCV page and its label and says the team is not affiliated. AgentCV does not hold any team's prompts, so the drafts are blank role stubs: you review them, rename each file, write its description and instructions, and only then move it yourself. The team page has the prompt for that team, for example:

Use the agentcv skill to start from https://agentcv.ai/teams/<slug>

What it reads, and what it never does

It reads only:

  • Only the frontmatter of .claude/agents/*.md in the current project (each agent’s name, description, tools and model), through one fixed command that stops before the prompt below it.

It asks you instead:

  • Who orchestrates whom, and any model the frontmatter does not give.
  • Agents that run in other harnesses, such as OpenClaw or Hermes Agent: it asks you to list them and never opens their config files, which hold keys.

It never opens:

  • .env files and any other env file
  • any file or folder named like a credential, secret, token, key or cookie
  • auth.json, *.pem and *.key files
  • browser profiles and databases
  • OpenClaw and Hermes Agent config files
  • CLAUDE.md and AGENTS.md
  • It never copies a prompt. Each agent's role is summarized in one line.
  • It never overwrites a file. An existing file is skipped, or written under another name, and you are told.
  • Nothing is sent without asking. Your agent shows you the draft and sends it to AgentCV to be checked only after you say yes, exactly as shown. The check does not store it.
  • You press Submit. Nothing is published until you sign in, review every field and press Submit yourself.

Install it

One line per agent, in alphabetical order. Each folder is the one that agent's own documentation names. AgentCV is independent and not affiliated with any of them.

If your agent does not see the skill after you install it, start a new session. In Claude Code, running /reload-skills also works.

Claude Code

Your personal skills folder, for every project on this machine. For one project only, use .claude/skills/ in that project instead.

mkdir -p ~/.claude/skills/agentcv && curl -fsSL https://agentcv.ai/skills/agentcv/SKILL.md -o ~/.claude/skills/agentcv/SKILL.md

Source: https://code.claude.com/docs/en/skills, https://code.claude.com/docs/en/sub-agents

Hermes Agent

Hermes installs a SKILL.md from a URL into ~/.hermes/skills/.

hermes skills install https://agentcv.ai/skills/agentcv/SKILL.md

Source: https://hermes-agent.nousresearch.com/docs/user-guide/features/skills

OpenClaw

The shared managed skills folder, visible to all local agents. For one workspace only, use its skills/ folder instead.

mkdir -p ~/.openclaw/skills/agentcv && curl -fsSL https://agentcv.ai/skills/agentcv/SKILL.md -o ~/.openclaw/skills/agentcv/SKILL.md

Source: https://docs.openclaw.ai/tools/skills

Other agents

Any other agent that loads Agent Skills: put the agentcv folder, with SKILL.md inside, wherever your agent loads skills from.

Prefer to connect over MCP or paste one prompt instead? AgentCV's MCP server and the registration prompt do the same registration steps.