Guide · Agent skill
How to use the Launch Legal Review agent skill (launch-review)
Updated · written with AI from the project's own documentation and checked against our catalogue data
The launch-review skill is a reusable instruction package for an AI agent. When you ask it to review a product launch or whether a feature can ship, it reads the product requirements, checks the launch against each category of your legal review framework, and writes a review memo that says what has to happen before release. This guide is for in-house legal and product-legal teams, and for product managers who need to know which legal steps block a launch. By the end you will know what the skill needs, how to install it, how to trigger it, and where its limits are.
Launch Legal Reviewanthropics/claude-for-legal · Apache-2.0 Download the skill Quick start
- 1Confirm the plugin configuration. Make sure ~/.claude/plugins/config/claude-for-legal/product-legal/CLAUDE.md exists and has no placeholders in its framework and calibration sections.
- 2Download the skill ZIP. Download the launch-review ZIP from this page.
- 3Install the skill. Unzip it into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
- 4Gather the launch materials. Collect the PRD, any spec or design doc, the marketing plan if there is one, the launch date, and the tracker ticket ID if you use one.
- 5Ask for the review. Say something like 'review this launch' or run /product-legal:launch-review with the PRD file, Drive link, or ticket ID.
- 6Read the memo and follow up. Check the blockers and informational items, verify flagged citations, and follow any handoffs to related review skills.
What the launch-review skill makes the agent do
An agent skill is a folder of instructions (a SKILL.md file plus reference files) that an AI agent loads when a request matches the skill's description. This one covers launch reviews. The agent reads the PRD and related documents, walks every category of the review framework, calibrates each finding against your team's risk table, and outputs a review memo in your house format.
The skill is designed to use your team's own calibration file, so findings reflect what actually blocks launches at your company rather than a generic checklist. It also checks for AI components, even when the PRD does not label them as AI, and offers routing to related review skills. The skill itself states that a lawyer decides whether to accept lower-confidence sources.
- Reads the PRD, spec, marketing plan and launch date, and pulls the launch tracker ticket if a tracker is connected.
- Checks each framework category: contractual commitments, privacy, security, IP, third-party, and regulatory.
- Flags any AI component and runs the AI use-case triage skill alongside the framework walk.
- Tags citations by source and confidence tier, and flags novel issues explicitly.
- Writes a memo stating what must happen before shipping, and can post a summary to the ticket if connected.
Set up the framework and calibration file first
The skill expects the claude-for-legal product-legal plugin to be configured. It loads a CLAUDE.md file at ~/.claude/plugins/config/claude-for-legal/product-legal/CLAUDE.md and reads its Review framework, Risk calibration, Launch review process and Escalation sections.
If that file still contains placeholders, the skill is designed to stop. A default eight-category framework exists as a fallback, but the skill notes that it may not match local practice, so teams should fill in their own configuration before relying on the output.
- Review framework: the categories the agent checks.
- Risk calibration: what blocks a launch versus what is informational at your company.
- Launch review process: the output format for the memo.
- Escalation: when to route an issue up.
~/.claude/plugins/config/claude-for-legal/product-legal/CLAUDE.mdHow to install the skill
Download the skill ZIP from this page. You can install it in one of two ways. For a personal Claude Code setup, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder. Both locations are described in the source instructions for this skill.
If you use the Claude apps instead, upload the ZIP under Settings → Capabilities → Skills.
- Personal use: unzip into ~/.claude/skills/.
- Project use: unzip into the project's .claude/skills/ folder.
- Claude apps: Settings → Capabilities → Skills, then upload the ZIP.
How to trigger the skill with natural requests
You do not need to memorise a command. The skill's description lists phrases that should activate it, so ordinary requests work. You can also invoke it directly as a slash command with a PRD file, a Drive link, or a tracker ticket ID as the argument.
Be specific about the feature or product. The skill works best when it has the PRD or ticket in hand, so include the file or link in the same request.
- 'Review this launch' with the PRD attached.
- 'Legal review for [feature]'.
- 'Can we ship this?'
- 'What are the legal issues with [product]?'
- A request that references a launch tracker ticket or PRD needing a category-by-category memo.
/product-legal:launch-review PROJ-1234Example: a feature review from start to memo
Suppose a PM shares a PRD for a feature that collects new user data and adds a personalised recommendation step, then asks whether it can ship. The agent first checks for AI components, because words like 'personalized' and 'suggested' are treated as signals. It then walks each framework category, such as privacy for the new data collection and third-party for any new vendor.
Each finding is calibrated against your team's table. If your table says a particular kind of data collection needs a quick privacy assessment that ships in one to two days, the memo should reflect that practice rather than a heavier process. The output is a memo that tells the PM which steps must happen before release.
- Expect the memo to name blockers separately from informational items.
- Novel issues are flagged explicitly.
- If marketing copy is substantial, the skill hands off to the marketing-claims-review skill.
- If one finding needs more depth, the skill hands off to feature-risk-assessment.
Safety, network use and privilege
The skill does not run scripts. It does use the network: it can pull tickets from a connected Jira or Linear tool and run legal research through a connected tool such as Westlaw or CourtListener. These connectors are optional, and the source lists them as such.
The skill includes a destination check. If you name a destination such as a public channel, a company-wide list, a counterparty, or a vendor, it asks whether that destination is inside the privilege circle and offers a privileged version, a sanitised version, or both. It does not silently apply a privileged header and then help you paste the output somewhere the header will not protect it. Choose the destination carefully regardless.
- Runs scripts: no.
- Uses network: yes, for optional connectors and research tools.
- Public or company-wide channels can waive privilege. Check before sharing.
- If legal research returns thin results, the skill reports this and asks how to proceed. It does not silently fill gaps from web search or model memory.
Tips for getting reliable output
Treat the memo as a first-pass draft for lawyer review. The skill tags citations as [settled], [verify] or [verify-pinpoint], and tool-retrieved citations keep their source tag. Pinpoint citations carry the highest fabrication risk and should always be checked against a primary source.
Connect the launch tracker if you have one, because earlier ticket comments often hold context the PRD leaves out. Keep the calibration file current so findings match current practice.
- Give the skill the launch date so it can calibrate urgency.
- Check every [verify-pinpoint] citation against a primary source before relying on it.
- Use the skill for consistent first-pass reviews, not final sign-off on high-stakes or novel regulatory questions.
- Do not use it if your team has no configured framework or calibration file.
What people use it for
- A product manager asks which legal steps block a new feature before the launch date.
- An in-house lawyer runs a consistent first-pass review across several launches in a quarter.
- A team checks whether a new third-party integration or data collection needs routing to a further review.
- A PM reviews whether substantial marketing claims need a separate marketing-claims review before release.
Frequently asked questions
Is the launch-review skill free?
The source for this page lists the licence as Apache-2.0 but does not state pricing, so check the listing for any cost details. Optional connectors such as Jira, Linear, Westlaw or CourtListener may have their own terms.
What license does the launch-review skill use?
The skill is listed under the Apache-2.0 licence. Read the licence text in the download before reusing or redistributing it.
What do I need before the launch-review skill works?
You need the claude-for-legal product-legal plugin configured, with a CLAUDE.md file containing your review framework and risk calibration. The Jira or Linear connector and a legal research tool are optional.
Why does the skill stop and say my setup is incomplete?
The skill is designed to stop when placeholders are present in your CLAUDE.md. Fill in your framework and calibration sections so the review matches your team's practice.
Can the skill replace a lawyer's review?
No. The output is a draft for lawyer review, and the skill itself says a lawyer decides whether to accept lower-confidence sources. It is not a substitute for final sign-off on high-stakes or novel regulatory questions.
What should I use instead if I have no configured framework?
The source says the default fallback is a generic eight-category list that may not match local practice. Configure your own framework first, or use a human-led review process until you have one.
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
This skill suits in-house legal and product-legal teams that have configured the claude-for-legal product-legal plugin and want a consistent first-pass launch review. Product managers who need to know which legal steps block a launch will also find it useful. Teams without a framework or calibration file, or those needing final sign-off on high-stakes regulatory questions, should choose a different approach.
Download the skill