How to use the Cold-Start Interview Setup agent skill

What the skill makes the agent do

Every regulator publishes constantly, and most of it will not matter to you. This skill runs a setup interview that learns which regulators to watch, where your policies are kept and what 'material' means for your work. It then writes those answers to a configuration file that the regulatory-legal plugin reads later, so its monitoring and digests filter for what matters to you.

The skill can also re-check which connected tools actually respond, without re-running the whole interview. Its job is configuration and connection checks only. The skill itself is not a source of legal advice, and its output still needs professional review.

• Asks which regulators you follow and what counts as material to you
• Connects and indexes your policy folder
• Writes a completed configuration file to ~/.claude/plugins/config/claude-for-legal/regulatory-legal/CLAUDE.md
• Re-checks connectors on request with --check-integrations

What you need before installing

The skill depends on the regulatory-legal plugin from the claude-for-legal project. You also need local file access to the ~/.claude plugin configuration folder, since that is where the configuration is written.

Connectors or MCP tools are optional. If you want the skill to verify them, you must set them up separately first.

• The claude-for-legal regulatory-legal plugin installed
• Local file access to the ~/.claude plugin configuration folder
• Any connectors or MCP tools you want verified (optional, set up separately)

How to install the skill

Download the ZIP for this skill from the page where you found this guide. Then choose one of two routes. For a personal install available in all your projects, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload the skill

How to trigger the interview with natural requests

The skill is named cold-start-interview and runs as /cold-start-interview. On a fresh install, you can ask the agent in plain words to set up the regulatory-legal plugin or start the cold-start interview, and it should pick up the skill from its description.

The skill's description lists three situations where it applies: a fresh install, reconfiguring with --redo, and re-checking connectors with --check-integrations. Use the flag that matches what you need.

• Fresh install: ask the agent to start the cold-start interview for regulatory-legal
• Change settings later: use /cold-start-interview --redo
• Check connectors only: use /cold-start-interview --check-integrations

What happens during the interview

The skill first checks whether a configuration file already exists. If one is populated, it stops and asks you to use --redo to change it. If a previous setup was paused, it offers to resume from where it stopped.

It begins with a short preamble that offers a quick path of about 2 minutes or a full path of about 15 minutes. The full path adds your full watchlist, materiality thresholds, feed cadence, policy library index and comment-period sources. You can upgrade from quick to full later.

If a shared company profile exists at ~/.claude/plugins/config/claude-for-legal/company-profile.md, the skill reads it and asks you to confirm rather than re-asking company questions. Once the interview finishes, the skill writes the configuration file, creating parent folders as needed.

• Quick path: about 2 minutes for role, practice setting and primary regulatory regime
• Full path: about 15 minutes, adding watchlist, thresholds, feed cadence and policy index
• The file it writes: ~/.claude/plugins/config/claude-for-legal/regulatory-legal/CLAUDE.md

Re-check which connectors actually respond

Use the --check-integrations flag when you want to know which connected tools work. The skill skips the interview and re-runs only the connection check, then rewrites the Available integrations table in your configuration file.

A connector is marked ✓ only if a tool call actually succeeded. A connector that is configured but not yet tested is marked ⚪, with a one-line note on how to confirm it. The skill does not mark a connector ✓ just because it is declared in a configuration file, since that would suggest it works when it may not.

• ✓ means a tool call succeeded during the check
• ⚪ means configured but not tested yet, with a how-to for confirming

Example requests you can make

You can describe what you want in ordinary language. Examples include asking to set up the regulatory watchlist for a new install, asking to change your materiality threshold, or asking whether your connected policy sources are responding.

If you are not doing regulatory or compliance work, the skill's preamble says the plugin is aimed at people who track regulatory developments and manage compliance obligations, and it points you to /legal-builder-hub:related-skills-surfacer.

• Set up my regulatory watchlist for the first time
• Redo my setup; I want to change which regulators I follow
• Check whether my connected tools are responding before I rely on the monitor

Safety notes: scripts and network

This skill runs no scripts and uses no network access. Its work is a guided interview, indexing of documents you point it at, and writing a plain-text configuration file.

The skill is designed not to use your personal material. Its instructions explicitly forbid reading your personal Claude history or home-folder memory files, such as ~/CLAUDE.md or ~/user.md, to fill in the interview. The only inputs are your typed answers and documents you point at or paste in. If something relevant came up earlier in the conversation, the skill should ask before using it.

• Runs scripts: no
• Uses network: no
• Does not read personal Claude history or home-directory memory files
• Only uses your typed answers and documents you provide

Tips for a useful setup

The more specific your answers, the quieter and more useful the digests will be. A generic configuration gives generic output, so spend the time on what 'material' means to you. You can paste a link or a document for anything longer than a sentence instead of typing it from memory.

Connect the policy folder you want indexed. If you install user-scoped rather than project-scoped, the skill can read documents outside a single project folder, such as Downloads or Documents.

• Describe your materiality threshold in concrete terms, not general ones
• Paste a document for the company description or policy library rather than retyping it
• Run --check-integrations after connecting any new tool