How to use the reg-gap-analysis agent skill for AI law gaps

What does the reg-gap-analysis skill make the agent do?

The skill runs a fixed workflow. The agent first reads your governance file at ~/.claude/plugins/config/claude-for-legal/ai-governance-legal/CLAUDE.md to confirm that the regulatory footprint and use case registry are filled in. It then decides whether the regulation applies, looking at jurisdiction, thresholds, sector carve-outs and whether your company is a builder or a deployer. If the rule does not apply, the agent says so in one line and stops.

If it applies, the agent extracts each substantive requirement, compares it with your current state (use case registry, vendor positions, AI policy commitments and impact assessment practices), and rates each item as a gap. The output sorts items into must-do, should-do, already compliant, or accepted gaps, each with an effort estimate, and saves a dated markdown document for your file.

The skill is written to flag ambiguity. Where a regulation is unclear, the agent is instructed to say so instead of presenting an uncertain point as settled.

How do I install the reg-gap-analysis skill?

Download the skill ZIP from the directory page. Then install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into the project's .claude/skills/ folder. In the Claude apps, you can instead upload the ZIP under Settings, then Capabilities, then Skills.

The skill package contains a single file, SKILL.md. Keep the folder structure from the ZIP intact so the agent can find that file.

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

Prepare your CLAUDE.md before the first run

The analysis compares a new rule against what you already do, so the agent needs that information in a file it can read. The skill reads ~/.claude/plugins/config/claude-for-legal/ai-governance-legal/CLAUDE.md. The sections it looks for are Regulatory footprint, Use case registry, AI policy commitments, Vendor AI governance and Impact assessment house style.

If those sections are empty, the skill cannot produce a meaningful current-state comparison. Fill them in before you run it.

• Regulatory footprint: the jurisdictions and laws that already apply to you
• Use case registry: the AI systems you run and the conditions under which you run them
• AI policy commitments: what you have published or agreed to by contract
• Vendor AI governance: vendor positions already in place
• Impact assessment house style: your existing assessment practices

How do I trigger the skill with a natural request?

You can invoke the skill with its slash command, /reg-gap-analysis, and add the regulation name or pasted text as the argument. The skill's description also lists natural phrases that should load it, such as 'new reg just dropped', 'does [regulation] affect us', 'gap analysis for EU AI Act' and 'compliance check against [AI law or guidance]'. You can also paste regulatory text or attach a document.

The source shows the plugin-qualified form of the command as well, which is useful when several plugins are installed.

• Slash command: /reg-gap-analysis "EU AI Act high-risk provisions"
• Plugin-qualified form: /ai-governance-legal:reg-gap-analysis "EU AI Act high-risk provisions"
• Natural request: 'Does the Colorado AI law affect us?' with the regulatory text pasted or attached
• Natural request: 'Gap analysis for EU AI Act' against your CLAUDE.md

What needs to be in place to run it?

The skill needs Claude Code or a compatible agent that can read the local CLAUDE.md file. It also needs internet access or current regulatory sources, because the skill has the agent research the current regimes and verify that the law is still in force before it advises. The skill does not run scripts. It does use the network.

Treat the output as a first draft. The skill relies on the agent's own research of current law, which may be out of date or wrong. A qualified lawyer should review the findings before anyone acts on them.

• Claude Code or a compatible agent with access to the local CLAUDE.md file
• A populated CLAUDE.md with regulatory footprint, use case registry and policy sections
• Internet access or current regulatory sources for the research step
• Runs scripts: no. Uses network: yes.

Tips for getting useful output

Give the skill the specific regulation or the exact passage you care about. A narrow argument such as 'EU AI Act high-risk provisions' produces a more focused gap list than a broad request. Keep your use case registry current, since each gap is checked against it. Read the scope decision first; if the agent says the law does not apply, check the reason it gives against your footprint.

Use the accepted gaps section deliberately. It records risks you have chosen not to fix, along with the rationale and who accepted the risk. Keeping that record is more useful than leaving the gap unaddressed.

• Name a specific regulation or pasted passage rather than a general topic
• Keep the use case registry and vendor records current before each run
• Check the scoping decision against your builder or deployer role
• Use the dated markdown file as the working document for your legal team