Guide · Agent skill
How to use the Product Counsel Cold-Start Interview agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The cold-start-interview skill is a one-time setup interview for the product-legal plugin in the claude-for-legal project. It is for in-house product or privacy counsel who review launches and feature risk. After following this guide you will know what the skill does, how to install it, how to trigger it, and what it reads and writes.
Product Counsel Cold-Start Interviewanthropics/claude-for-legal · Apache-2.0 Download the skill Quick start
- 1Confirm the plugin is installed. Make sure the claude-for-legal product-legal plugin is installed, following its QUICKSTART.md. The skill cannot run without it.
- 2Install the skill folder. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills.
- 3Start the interview. Run /product-legal:cold-start-interview or ask the agent to set up product counsel.
- 4Choose quick or full. Pick quick for about 2 minutes or full for about 15 minutes. You can upgrade later with --full.
- 5Confirm the scope and company profile. If you are in a project folder, confirm project scope. If a shared company profile exists, confirm the one-line summary or say update.
- 6Review the calibration table. Let the agent read about 10 past launch reviews and check the blocked versus shipped table it builds. Correct anything that is wrong before it is saved.
- 7Check integrations. Run /product-legal:cold-start-interview --check-integrations after connecting tools, and look for ✓ marks that reflect real tool calls.
What is an agent skill, and what does this one do?
An agent skill is a folder containing a SKILL.md file that tells a Claude agent how to perform a specific task when you ask for it. This skill runs a setup interview for the product-legal plugin. It asks whether you want a quick (about 2 minutes) or full (about 15 minutes) setup.
It then reads about 10 of your past launch review documents from a launch tracker or cloud drive. From these it builds a table of what your company actually blocked versus shipped, and writes a practice profile file that later plugin commands use. The output is a configuration profile, not a legal opinion.
- Asks for a quick or full setup, then interviews you
- Reads about 10 past launch review documents from a tracker or Drive
- Builds a risk calibration table of blocked versus shipped decisions
- Writes your practice profile to the plugin config path
- Re-probes connected tools with --check-integrations, without re-interviewing you
Install the cold-start-interview skill
You can install the skill in one of two ways. Download the ZIP from the skill page on this directory and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder for that project only.
Alternatively, upload the ZIP in the Claude apps under Settings, then Capabilities, then Skills.
The skill assumes the claude-for-legal product-legal plugin is already installed. The source does not give full plugin installation steps; it points to a QUICKSTART.md file in the plugin for that.
- Option A: unzip into ~/.claude/skills/ (available everywhere on your machine)
- Option B: unzip into a project's .claude/skills/ folder (that project only)
- Option C: upload the ZIP in the Claude apps under Settings → Capabilities → Skills
How to trigger the skill with a request
You can run the skill as a slash command. The skill is also meant to be picked up from natural requests about first-time setup, onboarding product counsel, or a configuration that still has placeholders. The skill's own description lists these situations as triggers.
Two flags change its behaviour. --redo re-runs the full interview even if your profile is already populated. --check-integrations re-checks your connected tools only, and does not ask interview questions.
- Plain request: set up product counsel for the first time
- Plain request: my product-legal config still has placeholders, help me finish it
- Plain request: re-interview me for product counsel
- Plain request: check my product-legal integrations again
/product-legal:cold-start-interview
/product-legal:cold-start-interview --check-integrationsWhat happens during the interview
Before questions start, the skill checks whether a profile already exists and whether you are working in a project folder rather than your home directory. If you are in a project, it warns that it can only read files in that folder, and asks you to confirm the scope.
The skill then asks you to choose quick or full. Quick covers your role, review framework level (formal gate or advisory), and product or practice context. Full adds your risk calibration table, escalation matrix, review framework categories, house memo format, and launch tracker integration. You can upgrade later with --full.
If a shared company profile already exists at ~/.claude/plugins/config/claude-for-legal/company-profile.md, the skill shows a one-line summary and skips the company questions if you confirm it. The skill does not read your personal Claude history or your home-directory CLAUDE.md.
Connectors and the ✓ / ⚪ status
The skill can probe optional connectors: a launch tracker, document storage such as Google Drive, and Slack, via MCP. The source does not say in detail how each connector is set up.
A connector is marked ✓ only after a real MCP tool call succeeds. A connector that is configured but not tested is marked ⚪, with a one-line note on how to confirm it. Declarations in .mcp.json alone never count as working.
- ✓ means a real tool call succeeded
- ⚪ means configured but untested
- Use --check-integrations after connecting or disconnecting a tool
Where files are read and written
The skill writes your completed profile to ~/.claude/plugins/config/claude-for-legal/product-legal/CLAUDE.md, creating parent folders as needed. It shows the risk calibration table for your confirmation before finishing.
If a populated CLAUDE.md exists at the older cache path ~/.claude/plugins/cache/claude-for-legal/product-legal/*/CLAUDE.md but not at the config path, the skill copies it forward and shows you what moved.
- Profile written to ~/.claude/plugins/config/claude-for-legal/product-legal/CLAUDE.md
- Shared company profile at ~/.claude/plugins/config/claude-for-legal/company-profile.md
- Old cache copy migrated only if the config path is empty
What people use it for
- Setting up the product-legal plugin for the first time so launch reviews reflect your company's own past decisions
- Finishing a setup that was left mid-way, using the pause marker the skill resumes from
- Re-running the interview after your risk appetite or escalation contacts change, using --redo
- Confirming which connectors actually work after you connect or disconnect a launch tracker or Drive
Frequently asked questions
Is the cold-start-interview skill free?
The source does not state pricing or licence terms for the plugin beyond the listed licence of Apache-2.0 for this skill. Check the licence file in the repository for the terms that apply.
What do I need to run cold-start-interview?
You need Claude Code or a compatible agent with the claude-for-legal product-legal plugin installed. Launch tracker, document storage such as Google Drive, and Slack connectors via MCP are optional, and the skill probes them if present.
Does the skill read my documents?
Yes. It reads about 10 past launch review documents from your tracker or cloud drive to build the risk calibration table. If you do not want an AI agent reading those documents, do not run this skill.
How do I re-run the setup?
Use /product-legal:cold-start-interview --redo to re-interview from scratch. To only re-check connectors, use --check-integrations, which does not re-interview you.
Why does my connector show ⚪ instead of ✓?
A ⚪ means the connector is configured but no MCP tool call has succeeded yet. The skill only marks ✓ after a real tool call works, so run --check-integrations after confirming the connection.
Can the skill give me legal advice?
No. The skill creates a configuration profile that later plugin commands use. It is not a legal opinion, and it points users who do not review product launches to other tools.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use this skill if you are in-house product or privacy counsel setting up the claude-for-legal product-legal plugin and you are comfortable with an agent reading about 10 of your past launch reviews. Pick another approach if you do not review product launches or do not want those documents read by an AI agent, and do not rely on its output as legal advice.
Download the skill