Terms of Service Generator (AI Legal Assistant)
Scans any website or SaaS product and drafts complete, GDPR/CCPA-compliant Terms of Service with plain-English summaries for every section.
Narrow behavioral scope: WebFetch to the target URL and writing one Markdown file in the working directory; no credential access, destructive defaults, or covert exfiltration; [VERIFY] tags disclose assumptions and output includes a clear legal disclaimer. Deducted for: no user-confirmation step before writing, no rollback guidance, unverified publisher, unknown license.
Internally consistent instructions: trigger, three-step flow, and output template align; abnormal-input handling (mark undetermined items with VERIFY) is partially addressed. Deducted for: no tests or execution evidence (below the static cap of 10), unspecified behavior when WebFetch fails, and weak determinism since section depth depends on the model.
Trigger is precise (/legal terms <url>) and scope/non-fit rules exist (skip inapplicable sections like API terms for a blog). Deducted for: jurisdiction must be assumed and flagged VERIFY, output is English-only with no Chinese-language support, GDPR/CCPA always included which may over-trigger for non-EU/CA products.
Clear, well-layered document with guidelines and disclaimers. Deducted for: unknown license metadata, no versioning or changelog for this skill, no FAQ/known-limitations section, maintenance responsibility only inferable repo-wide, and a naming mismatch between the frontmatter name (terms-of-service-generator) and the directory (legal-terms).
Clear drafting-starting-point value: structured 16-section terms with plain-English summaries, a document-information table, and a [VERIFY] checklist, written directly to a usable Markdown file. Deducted for: static review cannot verify output quality; legal correctness requires attorney review (as the skill itself states); capped at 7 without third-party execution evidence.
Only the source files themselves are auditable; README statistics (82% of freelancers, 60-second reviews) are marketing claims not tied to this skill; no tests, sample outputs, or third-party corroboration. Static reviews cannot exceed 5, and evidence here is thin, so 3.
- Generated terms are not legal advice; a licensed attorney must review for the applicable jurisdiction before publication — the skill itself states this.
- GDPR/CCPA sections are unconditionally included and jurisdiction/contact details are largely [VERIFY] assumptions; direct publication carries compliance risk.
- The skill writes a file to the working directory without confirmation; output is English-only, so Chinese users must translate and localize themselves.
- Repository license unknown, publisher unverified, no versioning/changelog; maintenance and update path unclear.
What does this skill do, and when should you use it?
This is one of the 14 skills bundled in zubair-trabzada/ai-legal-claude, located at skills/legal-terms/SKILL.md and internally named terms-of-service-generator. Triggered by /legal terms <url>, it uses WebFetch to visit the target site, analyze its business type, data collection, accounts, and payment model, then drafts a complete Terms of Service covering up to 16 section types as a Markdown file. Every legal section comes with a Basecamp/Notion-style plain English summary, and anything the AI had to assume is tagged [VERIFY]. The output carries an explicit legal disclaimer stating it is a drafting starting point that must be reviewed by a licensed attorney, not legal advice.
Given a URL, it fetches the site with WebFetch and determines: business type (SaaS, marketplace, e-commerce, API service, etc.), core functionality, data collection (forms, cookies, payments), user accounts and user-generated content, API and third-party integrations, and target audience. It then drafts Terms of Service across sections such as acceptance, service description, accounts, payment terms, IP rights, UGC, prohibited uses, API terms, privacy with GDPR/CCPA subsections, disclaimers and liability limits, indemnification, termination, dispute resolution (including an arbitration opt-out), changes to terms, general provisions, and contact information. Sections that don't apply to the product are omitted. Output is written as TERMS-OF-SERVICE-[company]-[date].md in the working directory.
- An indie developer launching a new SaaS product who needs a first-draft ToS without immediately paying for legal drafting
- A small business standing up a website or storefront that wants GDPR/CCPA clauses included in an internal draft
- A web agency customizing terms to a client's actual business instead of reusing a generic template
- A creator or blogger with a simple content site who needs trimmed terms without irrelevant API or UGC sections
- An online service serving EU or California users that wants to confirm 'Do Not Sell' statements and data-subject rights are covered
What are this skill's strengths and limitations?
- Complete structure covering 16 section types, pruned to only the sections relevant to the actual product
- Plain English summaries per section, written in the accessible style of Basecamp/Notion policies
- Grounded in the real website's content rather than pure template filling
- GDPR and CCPA are always included as dedicated privacy subsections, with best practices like arbitration opt-out
- All assumptions tagged [VERIFY], making human review straightforward
- Output is a draft only — both the README and the output itself insist on review by a licensed attorney; not legal advice
- No license is stated for the repository, so adoption terms must be confirmed with the author
- No test suite, sample outputs, or user reviews to verify generation quality in advance
- Key items like governing law, refund policy, and contact details often need manual confirmation; automated assumptions may be wrong
- Relies on WebFetch to read the target site — results degrade for unreachable or sparse websites
- Running this skill requires installing the full 14-skill collection; no minimal install path is documented
How do you install this skill?
Install the whole skill collection: curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-legal-claude/main/install.sh | bash — one command installs all 14 skills, 5 agents, and the PDF scripts. Uninstall via the repo's uninstall.sh. The repo does not document installing this single skill in isolation; to do so you would need to manually place skills/legal-terms/ into your skills directory. The repository's license is unknown and not stated in the README.
How do you use this skill?
In Claude Code, run: /legal terms <url>, e.g. /legal terms https://example.com. The skill fetches the site, marks assumptions with [VERIFY], then writes TERMS-OF-SERVICE-[company]-[YYYY-MM-DD].md to your current working directory. Review every [VERIFY] item and have a licensed attorney check the result before publication. The repo does not say whether the skill can be invoked outside the slash-command flow.