[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M718AE0CCNEGEDRTKF0T97VK":3,"skill-same:01M718AE0CCNEGEDRTKF0T97VK":73},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":51,"commitSha":52,"skillDirectory":53,"forksCount":54,"repositoryHtmlUrl":55,"repositoryHomepageUrl":6,"ownerHtmlUrl":56,"ownerBlogUrl":57,"ownerDescription":6,"updatedAt":58,"install":59,"resources":72},"01M718AE0CCNEGEDRTKF0T97VK","policy-starter",null,"Draft a firm AI usage policy from published model policies, adapted to your practice profile — a research-and-synthesis tool whose output is a draft for attorney review and adoption, not a finished policy. Use when user says \"draft an AI policy\", \"we need an AI policy\", \"build an AI usage policy\", \"our firm needs a GenAI policy\", or similar requests to generate a first-cut internal AI policy.","anthropics\u002Fclaude-for-legal","anthropics","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F76263028?v=4","01M697XGDPXJAVQ8391V2BCYTE","Anthropic","01M7189QY7HXHMC9BW4MFHPZ6H",9977,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fai-governance-legal\u002Fskills\u002Fpolicy-starter\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000037","律师与法务","PROFESSION","适用职业",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"policy","starter","policy starter","use","tool","firm","for","usage","update","cut","hint","generate","genai","review","team","when","ai usage policy","only","model","scope","2026-11-10 10:12:30",false,"---\nname: policy-starter\ndescription: >\n  Draft a firm AI usage policy from published model policies, adapted to your\n  practice profile — a research-and-synthesis tool whose output is a draft for\n  attorney review and adoption, not a finished policy. Use when user says \"draft\n  an AI policy\", \"we need an AI policy\", \"build an AI usage policy\", \"our firm\n  needs a GenAI policy\", or similar requests to generate a first-cut internal\n  AI policy.\nargument-hint: \"[optional — scope hint, e.g. 'firm-wide', 'legal team only', 'update existing']\"\n---\n\n# \u002Fpolicy-starter\n\n1. Read `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002FCLAUDE.md`. If the practice profile is unpopulated, stop and direct to `\u002Fai-governance-legal:cold-start-interview`.\n2. Use the framework below.\n3. Run the scope interview — which sections does the policy need to cover, who's the audience, what's the deployment context. Do not skip to drafting.\n4. Web search for the current published model policies and guidance relevant to the deployment context (ABA, state bars, ILTA, CLOC, NIST, peer-firm \u002F peer-company policies, current state AI laws, EU AI Act, sector regulators as applicable).\n5. Draft the selected sections, sourced from the model policies, with `[review]` flags on every choice point and `[review]` open questions at the bottom of each section.\n6. Output with the draft header (\"DRAFT FOR INTERNAL LEGAL REVIEW — NOT FOR DISTRIBUTION\"), the sources block, the reviewer note, and the adoption checklist.\n7. Close with the next-steps decision tree.\n\n```\n\u002Fai-governance-legal:policy-starter\n\u002Fai-governance-legal:policy-starter \"we need an AI policy for our 30-lawyer firm\"\n\u002Fai-governance-legal:policy-starter \"update our existing policy for the 2026 state AI laws\"\n```\n\n---\n\n## Matter context\n\n**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 `\u002Fai-governance-legal:matter-workspace switch \u003Cslug>` or say `practice-level`.\" Load the active matter's `matter.md` for matter-specific context and overrides. Write outputs to the matter folder at `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002Fmatters\u002F\u003Cmatter-slug>\u002F`. Never read another matter's files unless `Cross-matter context` is `on`.\n\n---\n\n## Purpose\n\nA lot of firms and in-house teams don't have a written AI usage policy yet, or\nare running on a 2024-vintage one that doesn't mention the state AI laws, the EU\nAI Act implementing acts, the 2025 COPPA amendments, or what they actually ended\nup doing with Copilot and Claude for Work. This skill produces a **draft** policy\nto bring to the decision-maker — GC, managing partner, executive committee,\nboard, head of IT, head of HR — not a finished policy to circulate.\n\nThe discipline of this skill:\n\n1. **Source from published model policies, not from invention.** Search for and\n   read the ABA AI Toolkit, state bar guidance, ILTA's model policy, CLOC's\n   templates, and peer-firm \u002F peer-company policies that are public. Cite what\n   each source says and adapt it — don't generate policy language out of thin\n   air.\n2. **Decision-tree the scope before drafting.** A policy that tries to cover\n   everything covers nothing. Ask the user what sections the policy needs. Let\n   them pick. Then build each picked section with `[review]` flags on every\n   choice point.\n3. **Flag every judgment call.** The output is a draft the attorney reviews and\n   adopts; every threshold, every named tool, every disclosure trigger, every\n   enforcement consequence is a `[review]` line.\n4. **Header signals the scope of the audience.** This output may be read beyond\n   legal — by HR, IT, all staff. The header is adapted accordingly.\n\nThis skill does NOT finalize, distribute, publish, or even recommend a specific\nposition on the hard calls. It produces a draft and surfaces the choices.\n\n## Read `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002FCLAUDE.md` first\n\nBefore drafting, always read the practice profile. The sections that drive the\ndraft:\n\n- `## Company profile` — AI role (Builder \u002F Deployer \u002F Both), regulatory footprint,\n  external commitments, practice setting\n- `## Use case registry` — what's already approved, conditional, or a red line\n- `## AI policy commitments` — what a prior or current policy already says\n- `## Vendor AI governance` — what the team already requires from vendors\n- `## Governance team and escalation` — who approves, who escalates\n- `## Who's using this` — Role (lawyer \u002F non-lawyer) governs the header and the\n  \"adopt this\" framing\n\nIf `## AI policy commitments` is populated, this is an UPDATE, not a new draft —\ntreat the existing policy as the base and propose changes. If it's empty, this\nis a first-cut draft.\n\n## Scope interview (do this BEFORE drafting)\n\nAsk the user which sections the policy should cover. Present as a checklist —\nthe user picks, you build. Do not pre-decide.\n\n> **What should the AI policy cover? Pick the sections you want in the draft:**\n> 1. **Scope** — who the policy applies to (all staff, certain roles, contractors), what tools it covers (GenAI only, all AI, specific vendors), what data is in\u002Fout of scope.\n> 2. **Permitted and prohibited uses** — the approved categories, the red lines, the \"ask first\" cases.\n> 3. **Approval and review** — who approves a new tool, who approves a new use case, how the review request is filed, what the SLA is.\n> 4. **Disclosure** — to clients (for firms), to courts, to counterparties, to employees, to end users of an AI feature.\n> 5. **Data handling** — what confidential\u002Fclient\u002Fprivileged data can go where, data residency, vendor retention terms, training-on-data posture.\n> 6. **Training and certification** — who has to take training, on what cadence, consequences for non-completion.\n> 7. **Incidents and reporting** — what counts as an AI incident, how to report, who handles.\n> 8. **Enforcement** — what happens when the policy is violated, link to disciplinary framework.\n> 9. **Review cadence and ownership** — how often the policy gets updated, who owns updates, how changes are communicated.\n> 10. **Glossary** — defined terms (GenAI, approved tool, high-risk use, consequential decision, confidential data, etc.).\n>\n> Default starter pack for a firm \u002F in-house legal team that's never had a policy: 1, 2, 3, 4, 5, 9. Skip the rest for v1.\n\nAfter the user picks, ask the second question:\n\n> **Two more inputs before I draft:**\n> - **Audience** — who's reading this? (All staff \u002F legal team only \u002F attorneys plus staff \u002F client-facing version also needed) This drives tone and the glossary.\n> - **Deployment context** — (a) law firm, (b) in-house legal at a company (policy covers legal or company-wide?), (c) legal aid \u002F clinic, (d) government. This drives which model policies I search.\n\n## Source the model policies\n\nBefore drafting, run web searches for the most recent published model AI\npolicies and guidance.\n\n**Derive the model policy sources from the practice profile's `## Regulatory footprint`.** Don't hardcode US sources for a global user.\n\n| Jurisdiction | Model policy sources |\n|---|---|\n| US | ABA Formal Opinion 512, state bar guidance (CA, FL, NY, TX all have published AI guidance), ILTA model policy, CLOC templates, peer firm published AI policies |\n| UK | Solicitors Regulation Authority risk outlook, Law Society AI principles, ICO AI guidance, Bar Council guidance |\n| EU | EU AI Act compliance framework (Article 4 AI literacy, Article 17 quality management), national DPA AI guidance (CNIL, DSB, Garante, AEPD), EDPB guidelines, EU institutions' AI policies |\n| Australia | Law Council of Australia AI guidelines, OAIC AI guidance, state law society guidance, Australian AI Ethics Framework |\n| Singapore | PDPC Model AI Governance Framework, MinLaw guidance, MAS AI fairness principles (for financial services) |\n| Canada | Law Society of Ontario\u002FBC\u002FAlberta AI guidance, OPC AI guidance, TBS Directive on Automated Decision-Making |\n| Multi-jurisdiction | Use all applicable, and note where they diverge (e.g., EU requires human oversight documentation US doesn't; Australia focuses on voluntary ethics frameworks; Singapore focuses on sectoral regulation) |\n\nIf the practice profile's footprint is empty or `[PLACEHOLDER]`, ask: \"What jurisdiction(s) does your organization operate in? I'll draft from the model policies that match your regulatory environment and professional responsibility framework, not a US-centric template.\"\n\nFor each source the draft uses, **record it in a \"Sources\" block at the top of\nthe output** with: name, URL, date accessed, and what the draft took from it.\n\nIf a web search can't be run, note in the reviewer note: \"Could not run web\nsearch — draft sourced from training knowledge alone, verify against current\nversions of the cited sources before adopting.\" The verification log applies.\n\n## The draft\n\nOutput follows a consistent structure. **Every choice point gets a `[review]`\nflag.** The user has to decide; the skill presents options.\n\n### Header\n\n```\nDRAFT FOR INTERNAL LEGAL REVIEW — NOT FOR DISTRIBUTION\nPrepared for: [firm \u002F company name from practice profile]\nDate: [today's date]\nPrepared by: ai-governance-legal policy-starter skill, adapted from published model policies\nNot for adoption, distribution, posting, or reliance until reviewed, adapted, and approved by [attorney \u002F GC \u002F managing partner \u002F executive committee per the governance team section of the practice profile].\n```\n\nWhen the Role in `## Who's using this` is Non-lawyer: add a second line under\nthe header — \"If you are not a licensed attorney, solicitor, barrister, or other\nauthorised legal professional in your jurisdiction, bring this draft to your\nattorney contact ([name from practice profile]) before using any of it. This is\na starting draft for their review, not a policy you can adopt.\"\n\n### Sources block (at the top, under the header)\n\nA table of the model policies \u002F guidance \u002F regulations the draft drew from:\n\n| Source | URL | Accessed | What the draft took from it |\n|---|---|---|---|\n| ABA Formal Op. 512 | [url] | [date] | Disclosure and competence framing |\n| ILTA Model AI Policy v.[X] | [url] | [date] | Approval workflow, data handling |\n| [State] Bar Op. [X] | [url] | [date] | Disclosure to clients |\n| [peer firm] published AI policy | [url] | [date] | Scope language |\n| Colorado SB 24-205 | [url] | [date] | High-risk AI definition |\n| EU AI Act, Art. [X] | [url] | [date] | Vendor flow-down |\n\n### Executive summary\n\nThree paragraphs max. What the policy does, who it binds, what the reader has\nto do before it takes effect.\n\n### The sections\n\nOnly the sections the user picked, in the order above. For each:\n\n- A **header and scope** sentence.\n- The **substantive rules**, adapted from the cited model policies. Every\n  specific threshold, number, named tool, named vendor, or escalation contact\n  is `[review]`. Example: \"Confidential client data may not be entered into\n  [general-purpose consumer AI tools] `[review — list tools, or reference the\n  approved-tools list]`. Use of such data in [approved firm-licensed tools]\n  `[review — list tools]` is permitted subject to the data handling section.\"\n- **Source attribution** inline where a rule is adapted from a specific source.\n  Example: \"Attorneys must verify the accuracy of all AI-generated work product\n  before using it in representation of a client `[ABA Formal Op. 512]`.\"\n- **Open questions** at the bottom of each section — 2-3 decisions the attorney\n  needs to make before the section is ready. These are distinct from inline\n  `[review]` flags — these are the \"we don't have a position here yet\" items,\n  not the \"fill in the specifics\" items.\n\n### Adoption checklist\n\nAt the end of the draft, a checklist of the things that have to happen before\nthe policy is adopted. Don't invent these — pull from the practice profile's\ngovernance team and escalation section. Typical items:\n\n- [ ] Review by GC \u002F managing partner `[review — name]`\n- [ ] Review by IT \u002F security `[review — name]`\n- [ ] Review by HR (for enforcement \u002F training sections) `[review — name]`\n- [ ] Board \u002F executive committee approval (if required) `[review — confirm whether required]`\n- [ ] Training materials drafted\n- [ ] Announcement drafted\n- [ ] Effective date set `[review]`\n- [ ] Review cadence calendared `[review — annual is typical]`\n- [ ] Add policy to the `## AI policy commitments` section of the practice\n      profile once adopted\n\n### Reviewer note\n\nThe standard reviewer note above the header, per the `## Outputs` section of\nthe practice profile. Use the block format:\n\n> **⚠️ Reviewer note**\n> - **Sources:** web search ✓ \u002F not connected — cites from training knowledge\n> - **Read:** practice profile · [N] published model policies\n> - **Flagged for your judgment:** [N] `[review]` items inline · [N] open questions per section\n> - **Currency:** searched for developments since [date]\n> - **Before relying:** this is a DRAFT — bring to [approver from practice profile], don't distribute until adopted\n\n## Don'ts\n\n- **Don't invent policy language.** Every substantive rule in the draft must be\n  traceable to a cited source or flagged `[review — adapted, no direct source]`.\n- **Don't pick the hard calls for the attorney.** \"Should paralegals be\n  permitted to use AI for first-draft work?\" is a `[review]`, not a recommended\n  position.\n- **Don't produce a finished-looking policy.** The header, the reviewer note,\n  and the `[review]` flags throughout are the signal that this is a draft. Do\n  not soften them.\n- **Don't skip the scope interview.** If the user says \"just draft a full\n  policy,\" push back: \"A policy that tries to cover everything covers nothing.\n  Which sections do you want? Here's the checklist.\" One round of negotiation\n  is fine — two is also fine. Drafting without scope is the failure mode.\n- **Don't generate section content the user didn't ask for.** If they picked 1,\n  2, 3, 4, 5, 9, do those. Don't add section 6 because \"a real policy needs\n  training.\"\n- **Don't recommend a specific vendor, tool, or consequence.** Flag those\n  `[review]` with context on what a typical decision would be, not what the\n  user's should be.\n- **Don't promise legal sufficiency.** The draft is a starting point for\n  attorney review, not a tested policy.\n\n## Handoffs\n\nAfter the draft is produced, close with the decision tree from the practice\nprofile. The most common next steps:\n\n1. **Tune the draft** — the user walks through the `[review]` flags and resolves\n   them with the attorney; the skill re-runs with the decisions baked in.\n2. **Stakeholder summary** — produce a one-page version for the board or\n   executive committee explaining what the policy does and doesn't do.\n3. **Training materials** — once the policy is adopted, `\u002Fai-governance-legal:aia-generation` can be used to produce per-use-case training notes.\n4. **Vendor sweep** — once the policy is adopted, `\u002Fai-governance-legal:vendor-ai-review` should be run against the vendors the policy references to check conformance.\n5. **Gap check against new regulation** — pair with `\u002Fai-governance-legal:reg-gap-analysis` to test the draft against a specific regulation or guidance before adoption.\n\n## Output scope reminder\n\nThe document this skill produces reaches HR, IT, and the broader business — not\njust legal. Keep the language plain enough for non-lawyers to follow. The legal\nprecision is in the `[review]` flags and the sources, not in jargon.\n","# \u002Fpolicy-starter\n\n1. Read `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002FCLAUDE.md`. If the practice profile is unpopulated, stop and direct to `\u002Fai-governance-legal:cold-start-interview`.\n2. Use the framework below.\n3. Run the scope interview — which sections does the policy need to cover, who's the audience, what's the deployment context. Do not skip to drafting.\n4. Web search for the current published model policies and guidance relevant to the deployment context (ABA, state bars, ILTA, CLOC, NIST, peer-firm \u002F peer-company policies, current state AI laws, EU AI Act, sector regulators as applicable).\n5. Draft the selected sections, sourced from the model policies, with `[review]` flags on every choice point and `[review]` open questions at the bottom of each section.\n6. Output with the draft header (\"DRAFT FOR INTERNAL LEGAL REVIEW — NOT FOR DISTRIBUTION\"), the sources block, the reviewer note, and the adoption checklist.\n7. Close with the next-steps decision tree.\n\n```\n\u002Fai-governance-legal:policy-starter\n\u002Fai-governance-legal:policy-starter \"we need an AI policy for our 30-lawyer firm\"\n\u002Fai-governance-legal:policy-starter \"update our existing policy for the 2026 state AI laws\"\n```\n\n---\n\n## Matter context\n\n**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 `\u002Fai-governance-legal:matter-workspace switch \u003Cslug>` or say `practice-level`.\" Load the active matter's `matter.md` for matter-specific context and overrides. Write outputs to the matter folder at `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002Fmatters\u002F\u003Cmatter-slug>\u002F`. Never read another matter's files unless `Cross-matter context` is `on`.\n\n---\n\n## Purpose\n\nA lot of firms and in-house teams don't have a written AI usage policy yet, or\nare running on a 2024-vintage one that doesn't mention the state AI laws, the EU\nAI Act implementing acts, the 2025 COPPA amendments, or what they actually ended\nup doing with Copilot and Claude for Work. This skill produces a **draft** policy\nto bring to the decision-maker — GC, managing partner, executive committee,\nboard, head of IT, head of HR — not a finished policy to circulate.\n\nThe discipline of this skill:\n\n1. **Source from published model policies, not from invention.** Search for and\n   read the ABA AI Toolkit, state bar guidance, ILTA's model policy, CLOC's\n   templates, and peer-firm \u002F peer-company policies that are public. Cite what\n   each source says and adapt it — don't generate policy language out of thin\n   air.\n2. **Decision-tree the scope before drafting.** A policy that tries to cover\n   everything covers nothing. Ask the user what sections the policy needs. Let\n   them pick. Then build each picked section with `[review]` flags on every\n   choice point.\n3. **Flag every judgment call.** The output is a draft the attorney reviews and\n   adopts; every threshold, every named tool, every disclosure trigger, every\n   enforcement consequence is a `[review]` line.\n4. **Header signals the scope of the audience.** This output may be read beyond\n   legal — by HR, IT, all staff. The header is adapted accordingly.\n\nThis skill does NOT finalize, distribute, publish, or even recommend a specific\nposition on the hard calls. It produces a draft and surfaces the choices.\n\n## Read `~\u002F.claude\u002Fplugins\u002Fconfig\u002Fclaude-for-legal\u002Fai-governance-legal\u002FCLAUDE.md` first\n\nBefore drafting, always read the practice profile. The sections that drive the\ndraft:\n\n- `## Company profile` — AI role (Builder \u002F Deployer \u002F Both), regulatory footprint,\n  external commitments, practice setting\n- `## Use case registry` — what's already approved, conditional, or a red line\n- `## AI policy commitments` — what a prior or current policy already says\n- `## Vendor AI governance` — what the team already requires from vendors\n- `## Governance team and escalation` — who approves, who escalates\n- `## Who's using this` — Role (lawyer \u002F non-lawyer) governs the header and the\n  \"adopt this\" framing\n\nIf `## AI policy commitments` is populated, this is an UPDATE, not a new draft —\ntreat the existing policy as the base and propose changes. If it's empty, this\nis a first-cut draft.\n\n## Scope interview (do this BEFORE drafting)\n\nAsk the user which sections the policy should cover. Present as a checklist —\nthe user picks, you build. Do not pre-decide.\n\n> **What should the AI policy cover? Pick the sections you want in the draft:**\n> 1. **Scope** — who the policy applies to (all staff, certain roles, contractors), what tools it covers (GenAI only, all AI, specific vendors), what data is in\u002Fout of scope.\n> 2. **Permitted and prohibited uses** — the approved categories, the red lines, the \"ask first\" cases.\n> 3. **Approval and review** — who approves a new tool, who approves a new use case, how the review request is filed, what the SLA is.\n> 4. **Disclosure** — to clients (for firms), to courts, to counterparties, to employees, to end users of an AI feature.\n> 5. **Data handling** — what confidential\u002Fclient\u002Fprivileged data can go where, data residency, vendor retention terms, training-on-data posture.\n> 6. **Training and certification** — who has to take training, on what cadence, consequences for non-completion.\n> 7. **Incidents and reporting** — what counts as an AI incident, how to report, who handles.\n> 8. **Enforcement** — what happens when the policy is violated, link to disciplinary framework.\n> 9. **Review cadence and ownership** — how often the policy gets updated, who owns updates, how changes are communicated.\n> 10. **Glossary** — defined terms (GenAI, approved tool, high-risk use, consequential decision, confidential data, etc.).\n>\n> Default starter pack for a firm \u002F in-house legal team that's never had a policy: 1, 2, 3, 4, 5, 9. Skip the rest for v1.\n\nAfter the user picks, ask the second question:\n\n> **Two more inputs before I draft:**\n> - **Audience** — who's reading this? (All staff \u002F legal team only \u002F attorneys plus staff \u002F client-facing version also needed) This drives tone and the glossary.\n> - **Deployment context** — (a) law firm, (b) in-house legal at a company (policy covers legal or company-wide?), (c) legal aid \u002F clinic, (d) government. This drives which model policies I search.\n\n## Source the model policies\n\nBefore drafting, run web searches for the most recent published model AI\npolicies and guidance.\n\n**Derive the model policy sources from the practice profile's `## Regulatory footprint`.** Don't hardcode US sources for a global user.\n\n| Jurisdiction | Model policy sources |\n|---|---|\n| US | ABA Formal Opinion 512, state bar guidance (CA, FL, NY, TX all have published AI guidance), ILTA model policy, CLOC templates, peer firm published AI policies |\n| UK | Solicitors Regulation Authority risk outlook, Law Society AI principles, ICO AI guidance, Bar Council guidance |\n| EU | EU AI Act compliance framework (Article 4 AI literacy, Article 17 quality management), national DPA AI guidance (CNIL, DSB, Garante, AEPD), EDPB guidelines, EU institutions' AI policies |\n| Australia | Law Council of Australia AI guidelines, OAIC AI guidance, state law society guidance, Australian AI Ethics Framework |\n| Singapore | PDPC Model AI Governance Framework, MinLaw guidance, MAS AI fairness principles (for financial services) |\n| Canada | Law Society of Ontario\u002FBC\u002FAlberta AI guidance, OPC AI guidance, TBS Directive on Automated Decision-Making |\n| Multi-jurisdiction | Use all applicable, and note where they diverge (e.g., EU requires human oversight documentation US doesn't; Australia focuses on voluntary ethics frameworks; Singapore focuses on sectoral regulation) |\n\nIf the practice profile's footprint is empty or `[PLACEHOLDER]`, ask: \"What jurisdiction(s) does your organization operate in? I'll draft from the model policies that match your regulatory environment and professional responsibility framework, not a US-centric template.\"\n\nFor each source the draft uses, **record it in a \"Sources\" block at the top of\nthe output** with: name, URL, date accessed, and what the draft took from it.\n\nIf a web search can't be run, note in the reviewer note: \"Could not run web\nsearch — draft sourced from training knowledge alone, verify against current\nversions of the cited sources before adopting.\" The verification log applies.\n\n## The draft\n\nOutput follows a consistent structure. **Every choice point gets a `[review]`\nflag.** The user has to decide; the skill presents options.\n\n### Header\n\n```\nDRAFT FOR INTERNAL LEGAL REVIEW — NOT FOR DISTRIBUTION\nPrepared for: [firm \u002F company name from practice profile]\nDate: [today's date]\nPrepared by: ai-governance-legal policy-starter skill, adapted from published model policies\nNot for adoption, distribution, posting, or reliance until reviewed, adapted, and approved by [attorney \u002F GC \u002F managing partner \u002F executive committee per the governance team section of the practice profile].\n```\n\nWhen the Role in `## Who's using this` is Non-lawyer: add a second line under\nthe header — \"If you are not a licensed attorney, solicitor, barrister, or other\nauthorised legal professional in your jurisdiction, bring this draft to your\nattorney contact ([name from practice profile]) before using any of it. This is\na starting draft for their review, not a policy you can adopt.\"\n\n### Sources block (at the top, under the header)\n\nA table of the model policies \u002F guidance \u002F regulations the draft drew from:\n\n| Source | URL | Accessed | What the draft took from it |\n|---|---|---|---|\n| ABA Formal Op. 512 | [url] | [date] | Disclosure and competence framing |\n| ILTA Model AI Policy v.[X] | [url] | [date] | Approval workflow, data handling |\n| [State] Bar Op. [X] | [url] | [date] | Disclosure to clients |\n| [peer firm] published AI policy | [url] | [date] | Scope language |\n| Colorado SB 24-205 | [url] | [date] | High-risk AI definition |\n| EU AI Act, Art. [X] | [url] | [date] | Vendor flow-down |\n\n### Executive summary\n\nThree paragraphs max. What the policy does, who it binds, what the reader has\nto do before it takes effect.\n\n### The sections\n\nOnly the sections the user picked, in the order above. For each:\n\n- A **header and scope** sentence.\n- The **substantive rules**, adapted from the cited model policies. Every\n  specific threshold, number, named tool, named vendor, or escalation contact\n  is `[review]`. Example: \"Confidential client data may not be entered into\n  [general-purpose consumer AI tools] `[review — list tools, or reference the\n  approved-tools list]`. Use of such data in [approved firm-licensed tools]\n  `[review — list tools]` is permitted subject to the data handling section.\"\n- **Source attribution** inline where a rule is adapted from a specific source.\n  Example: \"Attorneys must verify the accuracy of all AI-generated work product\n  before using it in representation of a client `[ABA Formal Op. 512]`.\"\n- **Open questions** at the bottom of each section — 2-3 decisions the attorney\n  needs to make before the section is ready. These are distinct from inline\n  `[review]` flags — these are the \"we don't have a position here yet\" items,\n  not the \"fill in the specifics\" items.\n\n### Adoption checklist\n\nAt the end of the draft, a checklist of the things that have to happen before\nthe policy is adopted. Don't invent these — pull from the practice profile's\ngovernance team and escalation section. Typical items:\n\n- [ ] Review by GC \u002F managing partner `[review — name]`\n- [ ] Review by IT \u002F security `[review — name]`\n- [ ] Review by HR (for enforcement \u002F training sections) `[review — name]`\n- [ ] Board \u002F executive committee approval (if required) `[review — confirm whether required]`\n- [ ] Training materials drafted\n- [ ] Announcement drafted\n- [ ] Effective date set `[review]`\n- [ ] Review cadence calendared `[review — annual is typical]`\n- [ ] Add policy to the `## AI policy commitments` section of the practice\n      profile once adopted\n\n### Reviewer note\n\nThe standard reviewer note above the header, per the `## Outputs` section of\nthe practice profile. Use the block format:\n\n> **⚠️ Reviewer note**\n> - **Sources:** web search ✓ \u002F not connected — cites from training knowledge\n> - **Read:** practice profile · [N] published model policies\n> - **Flagged for your judgment:** [N] `[review]` items inline · [N] open questions per section\n> - **Currency:** searched for developments since [date]\n> - **Before relying:** this is a DRAFT — bring to [approver from practice profile], don't distribute until adopted\n\n## Don'ts\n\n- **Don't invent policy language.** Every substantive rule in the draft must be\n  traceable to a cited source or flagged `[review — adapted, no direct source]`.\n- **Don't pick the hard calls for the attorney.** \"Should paralegals be\n  permitted to use AI for first-draft work?\" is a `[review]`, not a recommended\n  position.\n- **Don't produce a finished-looking policy.** The header, the reviewer note,\n  and the `[review]` flags throughout are the signal that this is a draft. Do\n  not soften them.\n- **Don't skip the scope interview.** If the user says \"just draft a full\n  policy,\" push back: \"A policy that tries to cover everything covers nothing.\n  Which sections do you want? Here's the checklist.\" One round of negotiation\n  is fine — two is also fine. Drafting without scope is the failure mode.\n- **Don't generate section content the user didn't ask for.** If they picked 1,\n  2, 3, 4, 5, 9, do those. Don't add section 6 because \"a real policy needs\n  training.\"\n- **Don't recommend a specific vendor, tool, or consequence.** Flag those\n  `[review]` with context on what a typical decision would be, not what the\n  user's should be.\n- **Don't promise legal sufficiency.** The draft is a starting point for\n  attorney review, not a tested policy.\n\n## Handoffs\n\nAfter the draft is produced, close with the decision tree from the practice\nprofile. The most common next steps:\n\n1. **Tune the draft** — the user walks through the `[review]` flags and resolves\n   them with the attorney; the skill re-runs with the decisions baked in.\n2. **Stakeholder summary** — produce a one-page version for the board or\n   executive committee explaining what the policy does and doesn't do.\n3. **Training materials** — once the policy is adopted, `\u002Fai-governance-legal:aia-generation` can be used to produce per-use-case training notes.\n4. **Vendor sweep** — once the policy is adopted, `\u002Fai-governance-legal:vendor-ai-review` should be run against the vendors the policy references to check conformance.\n5. **Gap check against new regulation** — pair with `\u002Fai-governance-legal:reg-gap-analysis` to test the draft against a specific regulation or guidance before adoption.\n\n## Output scope reminder\n\nThe document this skill produces reaches HR, IT, and the broader business — not\njust legal. Keep the language plain enough for non-lawyers to follow. The legal\nprecision is in the `[review]` flags and the sources, not in jargon.\n",{"name":5,"description":49,"argument-hint":50},"Draft a firm AI usage policy from published model policies, adapted to your practice profile — a research-and-synthesis tool whose output is a draft for attorney review and adoption, not a finished policy. Use when user says \"draft an AI policy\", \"we need an AI policy\", \"build an AI usage policy\", \"our firm needs a GenAI policy\", or similar requests to generate a first-cut internal AI policy.\n","[optional — scope hint, e.g. 'firm-wide', 'legal team only', 'update existing']","main","4a6c651889c97cc9140580363c73e0eb17379c2b","ai-governance-legal\u002Fskills\u002Fpolicy-starter",1959,"https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal","https:\u002F\u002Fgithub.com\u002Fanthropics","https:\u002F\u002Fanthropic.com","2026-11-08 22:42:37",{"source":8,"skillName":5,"commands":60,"prompt":70,"zipAvailable":71},[61,64,67],{"tool":62,"command":63},"npx","npx skills add anthropics\u002Fclaude-for-legal --skill policy-starter",{"tool":65,"command":66},"bunx","bunx skills add anthropics\u002Fclaude-for-legal --skill policy-starter",{"tool":68,"command":69},"pnpm","pnpm dlx skills add anthropics\u002Fclaude-for-legal --skill policy-starter","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fai-governance-legal\u002Fskills\u002Fpolicy-starter\u002FSKILL.md",true,[],[74,102,122,141,163,176],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":85,"publishedAt":44},"01M718SG9G2R92ECCD0HYK2R5E","reg-feed-watcher","Check regulatory feeds now and report what's new since the last check, filtered by your materiality threshold. Use when the user says \"check the feeds\", \"what's new\", \"regulatory update\", when running from the scheduled agent, or when manually pasting a regulatory development for classification and diff.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Freg-feed-watcher\u002FSKILL.md",[80],{"categoryNo":81,"categoryName":82,"taxonomyType":83,"taxonomyTypeDesc":84},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[86,87,88,89,90,91,92,93,30,94,95,96,27,97,39,98,99,32,100,101],"reg","watcher","reg feed watcher","running","feeds","filtered","diff","development","classification","new","now","scheduled","check","last","report","regulatory",{"skillNo":103,"name":104,"nameZh":6,"description":105,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":106,"categories":107,"keywords":108,"publishedAt":44},"01M718SCRTBN55DMTVSPG67A7N","policy-redraft","Produce a proposed marked-up policy redraft that closes a gap found by \u002Fregulatory-legal:gaps or \u002Fregulatory-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.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Fpolicy-redraft\u002FSKILL.md",[],[24,109,92,30,110,111,112,27,113,39,114,115,37,116,117,118,119,120,101,121],"policy redraft","up","fix","produce","documents","drafting","gaps","draft","id","gap","mark","first","direct",{"skillNo":123,"name":124,"nameZh":6,"description":125,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":126,"categories":127,"keywords":128,"publishedAt":44},"01M718S8RV0C4NMVKG078CBFXP","policy-diff","Diff a specific regulatory change against the indexed policy library. Use when a reg has changed and you need to know which policies it touches and what the gap is, when the user says \"diff this reg against our policies\", \"which policy does this affect\", or \"gap analysis\", or when reg-feed-watcher hands off a material item.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Fpolicy-diff\u002FSKILL.md",[],[92,24,129,130,39,131,132,86,133,87,134,135,136,137,27,138,118,139,140,101],"policy diff","which","you","analysis","text","change","paste","library","material","gap analysis","what the","name",{"skillNo":142,"name":143,"nameZh":6,"description":144,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":145,"categories":146,"keywords":147,"publishedAt":44},"01M718S6521Y0FT4T24VCFKVG5","matter-workspace","Manage matter workspaces — create, list, switch, close, or detach the active matter (practice-level). Use when working across multiple clients or matters and you need to keep one engagement's context separate from another, or when a substantive skill needs to know which matter it's working in.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Fmatter-workspace\u002FSKILL.md",[],[148,149,130,150,151,152,95,27,39,131,153,154,155,156,157,158,159,160,161,162],"workspace","matter workspace","switch","manage","workspaces","keep","across","create","engagement","close","active","clients","list","it","context",{"skillNo":164,"name":115,"nameZh":6,"description":165,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":166,"categories":167,"keywords":168,"publishedAt":44},"01M718S3FPEJNQGCMQ1AR8M4HE","Open gaps tracker — what's flagged and not yet closed. Use when the user asks \"what gaps are open\", \"gap tracker\", \"remediation status\", or wants to close (--close GAP-ID) or risk-accept (--accept GAP-ID) a tracked gap.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Fgaps\u002FSKILL.md",[],[115,117,118,169,157,170,171,172,173,27,174,39,175],"risk","tracker","remediation","status","open","accept","user",{"skillNo":177,"name":178,"nameZh":6,"description":179,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":180,"categories":181,"keywords":183,"publishedAt":44},"01M718S06WSZ9Y24JKYM0VVPPE","gap-surfacer","Reference: shared gap- and comment-tracker framework backing \u002Fregulatory-legal:gaps and \u002Fregulatory-legal:comments. Tracks open policy gaps with remediation status, ingests gaps from policy-diff, surfaces what's open and aging, routes to owners, and notifies gap owners via Slack with per-send confirmation. Loaded by the gaps and comments skills before doing substantive work.","https:\u002F\u002Fgithub.com\u002Fanthropics\u002Fclaude-for-legal\u002Fblob\u002Fmain\u002Fregulatory-legal\u002Fskills\u002Fgap-surfacer\u002FSKILL.md",[182],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[118,184,185,186,92,24,187,173,188,189,190,191,115,192,193,194,195,170,171,196],"gap surfacer","shared","tracks","send","confirmation","aging","false","comments","with","work","slack","via","surfaces"]