ai-inventory
EU AI Act per-system inventory — track each AI system's role (provider, deployer, importer,…
Produce a proposed marked-up policy redraft that closes a gap found by /regulatory-legal:gaps or /regulatory-legal:policy-diff. A first draft for internal review — not for direct application to approved policy documents. Use when the user says "redraft the policy", "draft the
$ npx -y skills add anthropics/claude-for-legal --skill policy-redraft --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/policy-redraftContext preview
The summary Claude sees to decide when to auto-load this skill.
Produce a proposed marked-up policy redraft that closes a gap found by /regulatory-legal:gaps or /regulatory-legal:policy-diff. A first draft for internal review — not for direct application to approved policy documents. Use when the user says "redraft the policy", "draft the
name: policy-redraft description: Produce a proposed marked-up policy redraft that closes a gap found by /regulatory-legal:gaps or /regulatory-legal:policy-diff. A first draft for internal review — not for direct application to approved policy documents. Use when the user says "redraft the policy", "draft the policy fix", "mark up the policy", or when gap-surfacer hands off a gap for drafting. argument-hint: "[GAP-ID or gap description]"
1. Load `~/.claude/plugins/config/claude-for-legal/regulatory-legal/CLAUDE.md` → policy library index + practice profile. 2. Use the workflow below. 3. Gather inputs: the gap (from `/regulatory-legal:gaps` output or described directly), the current approved policy text, the rule text. 4. Verify the rule is current (per the policy-diff rule-status check). If you can't verify, emit the `⚠️ RULE STATUS UNVERIFIED` banner. 5. Produce a marked-up redraft of the affected policy section(s) — smallest-possible edit, `[verify]` tags carried through, inline comments explaining WHY each change was made. 6. Output a Policy Redraft Memo. Write it to a new file named `[policy-name]-proposed-redraft-[YYYY-MM-DD].md` — never write to the source policy document. 7. Do NOT close the gap in the tracker. The gap closes when the redraft is applied AND approved, which is the policy owner's action.
---
> This skill produces a **proposal**, not an edit. It writes to a new file with a clearly-marked draft filename. It never writes over a source policy document, and it never closes a gap in the tracker — the gap closes when the redraft is applied AND approved by the policy owner.
**Matter context.** Check `## Matter workspaces` in the practice-level CLAUDE.md. If `Enabled` is `✗` (the default for in-house users), skip the rest of this paragraph — skills use practice-level context and the matter machinery is invisible. If enabled and there is no active matter, ask: "Which matter is this for? Run `/regulatory-legal:matter-workspace switch <slug>` or say `practice-level`." Load the active matter's `matter.md` for matter-specific context and overrides. Write outputs to the matter folder at `~/.claude/plugins/config/claude-for-legal/regulatory-legal/matters/<matter-slug>/`. Never read another matter's files unless `Cross-matter context` is `on`.
---
Gap-surfacer finds the gap. Policy-diff names what needs to change. This skill takes the next step and produces a marked-up redraft of the affected policy section — small, specific, flagged — as a first draft for the policy owner's review.
These are the load-bearing rules. If any of them would be violated, stop and ask.
1. **This is a PROPOSAL, not an edit.** Never write directly to a source policy document. The output goes to a new file at `[policy-name]-proposed-redraft-[YYYY-MM-DD].md`, or into the matter workspace. Not `[policy-name].md`. 2. **Never close the gap in the tracker.** Gaps close when the redraft is APPLIED AND APPROVED — that is the policy owner's action, not yours. If the user says "close the gap now that you've redrafted it," decline: "I produce the proposal. The gap closes when you've reviewed, applied, and approved the change. When that's done, tell me and I'll update the tracker." 3. **"Apply this for me" is not in scope.** If the user asks you to apply the redraft to the source policy: "I don't apply policy changes — that's the policy owner's action after review and approval. I produce the proposal. When it's been reviewed and approved, tell me and I'll update the gap tracker." 4. **Confirm the policy version before redrafting.** If the user gives you a file, ask: "Is this the approved version of the policy, and is it the latest? A redraft against an outdated policy creates divergence." If they paste text, trust but flag in the reviewer note. 5. **Smallest-possible edit.** Strike a word before a sentence, a sentence before a paragraph, a paragraph before a section. Only touch sections affected by the gap. Don't restyle the policy. 6. **Carry `[verify]` tags through.** Any effective date, threshold, citation, or requirement that came from model knowledge or an unverified source gets tagged in the redraft itself, not just in the memo.
Three inputs are required. If any is missing, ask — don't infer.
One of:
One of:
One of:
If the rule text is partial or ambiguous, apply the **no silent supplement** rule from CLAUDE.md: offer the user the options (paste full text, point at primary source, web-search-with-verify-tag, or stop), and wait.
Use the same rule-status check pattern as `policy-diff`. Red flags that the rule may not be in force:
Reference agents, skills, and data connectors for the legal workflows we see most — in-house commercial, privacy, product, corporate, employment, litigation, regulatory, AI governance, IP, and the learning side of the practice (law school clinics and
EU AI Act per-system inventory — track each AI system's role (provider, deployer, importer,…
Run an AI impact assessment — structured intake, risk analysis, regulatory classification per…
Run the cold-start interview — learns your AI governance practice and writes…
Guided customization of your AI governance practice profile — change one thing without…
Manage matter workspaces — new, list, switch, close, or detach (practice-level).…
Keep the AI policy current with practice — weekly sweep of saved AIAs, triage results, and…