First workflows

After initialization, you do not need to memorize Agents Pack commands. Talk to Claude Code, Codex, or Cursor about the outcome you want — or run the same operations yourself with the CLI.

Every installation includes the required ap-manage-agents-pack skill. It teaches your coding agent to inspect the installation, preview changes, use the CLI as the source of truth, and verify the result. It must not work around lifecycle protections or edit generated provider copies directly.

For a fuller guide, see Manage via coding agent.

Add one MCP server to all coding agents

The required ap-add-mcp skill manages machine-level remote MCP configuration independently of project initialization:

Use ap-add-mcp to add the remote MCP server named docs at
https://example.com/mcp to Claude Code, Codex, and Cursor.
Preview the change before applying it, then verify every provider.

Or run the CLI directly:

agents-pack mcp add docs --url https://example.com/mcp --dry-run
agents-pack mcp add docs --url https://example.com/mcp --yes
agents-pack mcp status docs

Configuration and authentication are separate. See Manage MCP servers.

Check that everything is healthy

Ask:

Check my Agents Pack installation and explain whether everything is healthy.
Do not change anything.

Or run:

agents-pack status

Status is read-only. It reports scope, pack version, selected agents, components, and whether every managed output is clean, missing, modified, or malformed.

Browse what is available

Ask:

Show me which frontend-related components are available and recommend which
ones fit this project.

Or run:

agents-pack list
agents-pack list --available --kind skill
agents-pack list --installed

Browse the docs skill catalog and subagent catalog for descriptions.

Remember project knowledge across agents

Portable memory works automatically after initialization. Claude Code, Codex, and Cursor recall relevant repository memory and save verified durable learning as ordinary Markdown. Shared project knowledge is Git-trackable by default; user-, machine-, and checkout-specific memory stays ignored.

You can still be explicit:

Remember that releases are cut from main. Share this with the team.

What does this project remember about releases?

Memory cleanup is separate and runs only when you request it. See Portable project memory.

Keep project documentation current

Ask:

Use ap-refresh-repo-docs to inspect this feature and its current diff. Update
the canonical documentation for every material behavior or design change.
Follow the project's existing documentation structure. Keep important
decisions and their rationale with the feature or subsystem they govern, and
report what documentation changed or why none was needed.

The required core instructions already make living documentation part of feature completion. The skill provides the deeper maintenance and reconciliation workflow. See Keep project documentation current.

Install an official skill

Ask:

Install the ap-frontend-design skill. Preview the change first and wait for my
approval before applying it.

Or run:

agents-pack install ap-frontend-design --dry-run
agents-pack install ap-frontend-design --yes

See Install and remove components.

For React projects, see Develop with React for the two optional Vercel-derived skills. For executing a prepared plan through task implementers and independent reviews, see Subagent-driven development.

Preview several designs before choosing

Introduced in Core 0.31.0 and strengthened in Core 0.31.1, the Design Studio workflow builds genuinely distinct, shareable static HTML concepts before you decide what to integrate into your real product. Once your pack contains its optional components, ask:

Use ap-design-studio to build three distinct static HTML concepts for our
homepage. Keep our logo and copy, make the signup path obvious, and avoid the
rounded-card treatment from our previous attempts. Include mobile views and
package the pages so I can share them with the team before choosing.

See Design Studio for availability, setup, fresh critiques, iteration limits, and the individual skills you can run manually.

Check for content updates

Ask:

Check whether a new Agents Pack content version is available. Summarize the
release notes, but do not update anything yet.

Or run:

agents-pack update --check

CLI upgrades and content-pack updates are separate. Content updates do not replace the agents-pack executable. See Update, pin, and roll back.

Create a portable skill

Every installation also includes the required ap-create-new-skill skill. Ask:

Create a reusable Agents Pack skill named deploy-app. It should guide an agent
through safely deploying this project to staging or production, including
preflight checks and post-deployment verification.

Or scaffold with the CLI:

agents-pack create skill deploy-app \
  --description "Deploy this application safely. Use for staging or production deployment." \
  --yes

Edit the canonical source under .agents-pack/user/ (or ~/.agents-pack/user/ for global scope), then synchronize:

agents-pack sync --dry-run
agents-pack sync --yes
agents-pack status

Names you create cannot use the reserved ap- prefix. See Create user components.

Fork an official component

If an official skill is close but should become yours:

Fork ap-debug into a user-owned skill named my-debug, then show me which
canonical file I should customize.

Or:

agents-pack fork ap-debug --name my-debug --dry-run
agents-pack fork ap-debug --name my-debug --yes

See Fork and customize.

If the agent does not select the skill automatically

Name it explicitly:

Use ap-manage-agents-pack to check for available updates and explain what
would change. Do not apply the update.

Use ap-create-new-skill to create a portable skill for reviewing database
migrations before deployment.

These are skills, not special chat commands. The exact way your coding agent shows or invokes skills may differ, but the natural-language request can stay the same across Claude Code, Codex, and Cursor.

When to use the CLI directly

Running the CLI yourself is useful when:

• you want to see or control each exact command;
• you are scripting setup or maintenance;
• your coding agent is not available; or
• you are diagnosing a lifecycle or ownership problem.

Full command options are in the CLI reference. If something goes wrong, see Troubleshooting.