How to use the ECC Setup Wizard agent skill (configure-ecc)

What does the configure-ecc skill make the agent do?

An agent skill is a folder with a SKILL.md file that tells your coding assistant how to handle a specific task. When this skill is active, the agent runs an interactive setup wizard in the current harness (the coding assistant environment you are working in). It first inspects what is already installed, then asks only the questions that apply to that assistant.

The skill is written to avoid manual copying of plugins and cloning ECC into temporary folders. It shows a preview of the planned changes and asks for one yes or no before applying anything, then verifies the result.

• Installs, updates or reconfigures ECC in Claude Code, Codex or Kimi
• Keeps to the plugins, scopes and hook features each harness actually supports
• Shows a preview before any change is made

How do you install the configure-ecc skill?

Download the ZIP file for the configure-ecc skill from this page. Then install it in one of two ways.

Option one is a file-system install. Unzip the file into ~/.claude/skills/ for all your projects, or into the .claude/skills/ folder of a single project. Option two is an upload: in the Claude apps, open Settings, go to Capabilities, and upload the skill under Skills.

The skill is documented as a repair and reconfiguration path after ECC is in place. It does not replace the provider's built-in first-time install screen.

• File-system install: unzip into ~/.claude/skills/ (personal) or .claude/skills/ (project)
• App install: Settings, then Capabilities, then Skills, then upload

How do you trigger the skill with a natural request?

You do not need a special command. Describe what you want in plain language and the agent should pick up the skill when it matches the task. Requests such as asking the agent to install ECC, update it, or reconfigure its hooks fit the skill's purpose.

If the agent cannot tell which assistant it is running in, it is meant to show what it detected and ask you to confirm the target before running any change command.

• Example: help me install ECC in this project
• Example: reconfigure ECC and let me choose the hook mode
• Example: update my ECC setup and show me a preview first

What happens in the Claude Code wizard?

In Claude Code the wizard runs a fixed sequence. The agent first lists your current plugins and marketplace sources without changing anything. If ECC is already installed once, it treats the run as a reconfiguration. If it finds several ECC scopes, manual or old installs, broken settings or marketplace conflicts, it stops and shows the recovery steps it was given rather than guessing what to delete.

Next it asks two questions, once each. The first is the install scope: user applies to all your projects, project is shared through repository settings, and local is private to the current project. The second is the hook mode: off, minimal, standard or strict. off keeps the skills and commands but disables ECC hook automation, minimal enables only the lightest lifecycle and safety automation, standard balances quality and safety, and strict enables the most checks and reminders.

Then the agent runs a preview with dry-run, shows one summary of planned actions, and asks yes or no once. After a yes, it applies the same choices and checks that the exit code is 0 and that the scope and hooks in the result match what you picked. Finally it runs a plugin list to confirm the install and suggests /reload-plugins or a restart of Claude Code.

• Inventory: no changes made
• Two choices: scope (user, project or local) and hook mode (off, minimal, standard or strict)
• Preview with dry-run, then one yes or no
• Apply, verify, and suggest /reload-plugins or a restart

How do Codex and Kimi differ?

The skill branches by harness. In Codex it uses Codex's native plugin lifecycle and does not show Claude's scope or hook levels. Codex checks the marketplace and available plugins, adds the ECC marketplace if needed, and asks for confirmation before installing. Codex asks you to approve its own hook trust decision, and the skill does not claim that Claude's four hook modes map onto Codex.

In Kimi, ECC goes into the project surface at ./.kimi-code. The skill states that Kimi does not support the ECC Claude lifecycle hook profile, so it does not ask for Claude's scope or hook mode. It previews the install first, then uses a doctor check to confirm that instructions and skills stay inside ./.kimi-code.

• Codex: native plugin commands, no Claude scope or hook questions
• Kimi: project install to ./.kimi-code, hooks unsupported, no Claude hook questions
• Unknown harness: the agent shows its evidence and asks before changing anything

What does the skill run, and what should you check first?

This skill runs scripts and uses the network. It executes node and npx commands. It installs ECC from the marketplace at affaan-m/ECC, and npx fetches the public npm package ecc-universal. Before you use it, make sure you are comfortable with those scripts running and with network calls during setup.

Read the preview before you answer yes. The dry-run output lists the planned actions, marketplace actions and, for a scope change, the source and destination. If you decline, the skill is designed to stop without applying changes.

• Requires Claude Code, Codex or Kimi CLI with plugin support
• Requires Node.js and npx
• Requires access to the ECC marketplace (affaan-m/ECC) and network access
• Not suitable if you do not want scripts run or network calls made during setup

What tips help you get the best result?

Run the skill from the assistant you actually use day to day, since the flow depends on which harness it detects. If you move ECC from one scope to another, expect the wizard to explain the move and include the move flag in the commands it proposes. Choose off hook mode if you want the skills and commands without automation, and choose strict only if you want the most reminders.

Keep the verdict in mind: this is a reconfiguration tool. For a first install, start with the provider's own install screen and use this skill afterwards to adjust settings.

• Let the agent inspect first; the inventory step makes no changes
• Pick the hook mode deliberately: off, minimal, standard or strict
• For a scope change, read the migration line in the summary before saying yes