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-nda/SKILL.md with the internal name custom-nda-generator. Triggered by /legal nda <description>, it gathers the parties, NDA type (mutual/one-way/employee/vendor), purpose, and scope of confidential information, then drafts a full agreement with 13+ standard sections. Every legal clause is followed by a plain English annotation explaining its practical impact. Output is written to a local Markdown file (e.g. NDA-Acme-Beta-2025-01-01.md) with an embedded legal disclaimer. It produces a drafting starting point, not legal advice — the document must be reviewed by a licensed attorney before execution.
Reads the situation description entered after /legal nda; extracts or asks for missing critical information (full legal names, NDA type, purpose, types of confidential information); applies sensible defaults (2-year term, 3-year survival for business info, 5 years for trade secrets) and asks for jurisdiction if unspecified; drafts a complete agreement covering party identification, recitals, definition of confidential information, standard exclusions (public info, prior knowledge, independent development, third-party sources, legally compelled disclosure with notice obligation), receiving-party obligations, permitted disclosures, term and termination, survival, return or destruction of materials, breach remedies, no license/warranty, no obligation to proceed, governing law and dispute resolution, general provisions, and signature blocks; inserts a PLAIN ENGLISH annotation after each section; writes NDA-[Party1]-[Party2]-[date].md to the working directory with a key-terms quick reference table.
- A freelancer needs a one-way NDA before engaging a new client, protecting deliverables and source code
- Two companies discussing a potential partnership need a mutual NDA covering business plans, customer lists, and financial data
- A startup hiring an employee or contractor needs an employee NDA with employment-specific provisions (no guarantee of employment, at-will clarification, reasonable scope limits)
- A vendor or service provider needs access to company information under a vendor-form NDA
- A small business owner who needs a first-draft agreement quickly before handing it to a lawyer for review
How do you install this skill?
- Generated NDAs are not legal advice; have a licensed attorney review before execution — the output's own disclaimer states this explicitly.
- The skill is framed for common-law (primarily US state) contexts with no Chinese-language support; mainland-China users must separately validate jurisdiction fit, enforceability, and terminology.
- Repository license status is unknown; confirm licensing before commercial use.
- The repo's recommended one-line curl|bash installer was not executed or audited in this review; read the script before installing.
- This is a static review — no code was executed; actual legal quality, completeness, and abnormal-input behavior of generated text are independently unverified.
- Local filesystem
Claude Code CLI (for slash-command invocation)Python 3.8+ and reportlab only if used with the repo's PDF reporting skill
Install the whole collection: curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-legal-claude/main/install.sh | bash — this installs all 14 skills, 5 agents, and the PDF scripts in one command. The README does not document installing this single skill in isolation; manually, it presumably means copying the skills/legal-nda/ folder into your Claude Code skills directory, but the exact path is undocumented. Requires Claude Code with an active Anthropic API key.
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/zubair-trabzada/ai-legal-claude.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/legal-nda" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
In Claude Code, run /legal nda <description>, e.g. /legal nda mutual NDA between Acme Corp and Beta Inc for discussing a potential partnership or /legal nda one-way NDA for a freelance designer. If parties, type, purpose, or confidential-information scope are missing, the skill asks before proceeding; it also requires a jurisdiction before generating. The finished document lands in the current directory as NDA-[Party1]-[Party2]-[YYYY-MM-DD].md.
What are this skill's strengths and limitations?
- Structurally complete output covering every standard NDA section, including enforceability-critical standard exclusions and the compelled-disclosure notice obligation
- Plain English annotation after every clause makes practical impact understandable to non-lawyers
- Confidential-information definition is tailored per scenario (tech deals reference source code and algorithms; partnerships reference financials and customer lists)
- Prominent legal disclaimer embedded in every generated document, explicitly requiring attorney review
- Explicitly not legal advice — output must be reviewed by a licensed attorney before signing
- Cannot generate without a specified jurisdiction, and the document itself notes the template may not meet all jurisdiction-specific requirements
- Repository has no declared license (License: unknown) — confirm terms with the author before commercial use
- No test suite or legal-professional validation record; quality depends entirely on the model
- Output is Markdown only — no PDF or Word export (PDF reporting belongs to a sibling skill, not this one)
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| Custom NDA Generator this page | 47 · Use with care | ★ 1.8k | 7mo ago | — |
| Deep Risk Analysis (AI Legal Assistant Skill) | 42 · Not recommended | ★ 1.8k | 7mo ago | — |
| OfficeCLI Skill | 38 · Not recommended | ★ 32k | 13d ago | Apache-2.0 |
| NotebookLM Research Assistant Skill | 63 · Recommended | ★ 7.8k | 1mo ago | MIT |
| Software Copyright Materials Generator | 54 · Use with care | ★ 5.9k | 1mo ago | MIT |
How did FollowSkills review this skill?
The skill runs no scripts, requests no network access, touches no credentials, and writes only one .md file in the working directory — a small risk surface; output carries an explicit AI disclaimer and attorney-review prompt, which is responsible for a sensitive legal domain. Deducted for: no source attribution or citation for legal content; rollback depends on repo-level uninstall scripts rather than the skill itself; and the parent repo promotes a one-line curl|bash installer, a supply-chain trust gap (indirect, not this skill's own behavior).
Instructions are well-structured: explicit trigger, required/optional inputs with defaults, 15 mandated sections, annotation format, and output file naming — the happy path is statically inferable. Deducted for: full reliance on model-generated legal text with no script or determinism guarantees; no test evidence (static review caps at 10); handling of abnormal input is limited to a single 'ask the user' fallback with thin failure-feedback evidence.
Trigger is precise (/legal nda + description), variants (mutual/one-way/employee/vendor) and required/optional inputs are clearly separated, and NDA generation is a genuine common need. Deducted for: no declared capability boundaries or non-fit ranges (regulated industries, cross-border data agreements unaddressed); US-state-law framing with no Chinese-language support limits fit for FollowSkills' Chinese audience; only Markdown output format is specified.
SKILL.md is well organized: frontmatter, trigger, stepwise instructions, a full output template, and guideline sections with reasonable progressive disclosure and stable naming. Deducted for: unknown repository license; no version number or changelog for the skill; no worked examples or FAQ; unclear maintenance ownership and update path; the annotation format rests on convention rather than validation, a hidden assumption.
The goal — a customized, annotated NDA — is supported by a 15-section template, four variants, a key-terms reference table, mandatory jurisdiction elicitation, and direct Markdown file output, offering real marginal value over a blank template. Deducted for: static review cannot verify the legal quality or completeness ceiling of generated text; no representative sample output exists, so direct usability is inferred only — capped at 7.
Source files are complete and auditable; the clause checklist and template can be checked item by item, constituting auditable primary material. Deducted for: no tests, no CI coverage of this skill, no third-party execution evidence, and no real generated sample; README cost/success figures are marketing-style author claims not tied to this skill and lack independent corroboration — capped at 5.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →