Deep Risk Analysis (AI Legal Assistant Skill)
Clause-by-clause contract risk scoring (1-10), financial exposure estimates, and ready-to-use replacement clause language.
Skill scope is narrow: reads user-specified contract files/URLs, writes only one RISK-ANALYSIS.md file, no destructive defaults, and mandates a clear legal disclaimer in output. Deducted for: unknown license metadata, unverified publisher, no data-flow disclosure for URL fetching, and no rollback or overwrite-confirmation mechanism.
Instructions are self-consistent with clear steps and a complete output template; the happy path is plausible. Deducted for: no tests, no error handling for corrupt files, unreachable URLs or non-English contracts; 1-10 risk scoring is subjective with no reproducible rubric; failure feedback is thin.
Trigger is explicit (/legal risks <file>) and the scenario matches the described capability. Deducted for: no declared non-fit boundaries (non-contract text, non-common-law, non-English jurisdictions), no Chinese-language support, dependence on Anthropic API without mainland-China reachability assessment, and no file-existence validation.
SKILL.md is well structured with frontmatter and command documentation. Deducted for: unknown repository license, no versioning or changelog, no known-limitation disclosure (e.g., jurisdictional accuracy), no FAQ/troubleshooting; README is marketing-heavy ('60 seconds', unsourced $10,000+ figures), and maintenance ownership/update path is unclear.
The output template is professional and complete (risk matrix, hidden risks, priority list, replacement clause language), and drafting alternative clauses exceeds what casual manual reading yields. Deducted for: static review cannot verify output is directly usable; 'financial exposure estimates' risk producing pseudo-precise numbers requiring heavy rework; no representative output evidence exists.
Only author claims and README marketing figures (82%, 67%, $300-$500/hour) with no cited sources; no tests, no CI evidence, no third-party execution records; the scoring method has no external standard corroboration. The only auditable primary material is the SKILL.md itself.
- Output is AI-generated and cannot replace review by a licensed attorney; financial exposure estimates in particular should not be taken at face value.
- Fetching contracts via URL sends text to the Anthropic API — assess confidentiality obligations before using on sensitive documents.
- Risk scores are subjective model judgments; the same contract may yield different scores with no reproducible rubric.
- Repository license is unknown; confirm terms before commercial use.
- No Chinese-language support and dependence on the Anthropic API; mainland-China users should verify network reachability.
What does this skill do, and when should you use it?
This is one skill inside zubair-trabzada's ai-legal-claude repository (a collection of 14 skills), located at skills/legal-risks/SKILL.md and triggered via the /legal risks <file> command. It scores every contract clause for risk, actively hunts for commonly missed hidden risks such as definition landmines, cross-reference traps, and auto-renewal clauses, and for each high-risk clause provides the quoted text, a plain-English explanation, a financial exposure estimate, and balanced replacement language. Output is a structured RISK-ANALYSIS.md report with an embedded legal disclaimer. It is not a lawyer — the output is informational only, and a licensed attorney should still be consulted before signing.
Accepts a file path, URL, or pasted contract text; identifies the contract type (SaaS, employment, NDA, MSA, SOW, lease, vendor agreement), parties, effective date, and governing law. Scores each clause 1-10 across 11 risk categories (financial exposure, liability transfer, restrictive covenants, auto-renewal traps, IP assignment overreach, etc.). Specifically searches six hidden-risk patterns: definition-section landmines, cross-reference traps, buried carve-outs, survival clauses, incorporation by reference, and defined-term drift. For every clause scoring 5+, it outputs exact quoted text, risk category, score, risk indicator, plain-English translation, financial exposure estimate, and specific alternative clause language, then writes RISK-ANALYSIS.md containing a risk matrix, aggregated financial exposure, Top 5 fix-first priorities, and a risk distribution summary.
- A freelancer reviewing a client contract before signing to spot unfavorable terms and potential financial traps
- A small business owner checking a vendor or SaaS agreement for uncapped liability or auto-renewal pitfalls before execution
- An AI automation agency offering contract risk analysis as a paid service (the README suggests $500-$1,500 per review, though that pricing claim covers the whole collection's marketing pitch)
- A negotiator generating a prioritized fix list with replacement language to use in clause-by-clause redlining against opposing counsel
- Anyone holding an English contract who needs legal jargon translated into everyday language plus a worst-case dollar figure
What are this skill's strengths and limitations?
- Very concrete methodology: 11 explicit risk categories and 6 hidden-risk patterns rather than a vague 'review your contract' prompt
- Enforced output structure — quoted text + plain-English explanation + financial estimate + replacement clause — makes every finding verifiable
- Explicitly instructs not to inflate risk: if a clause is fine, say so; review from the non-drafting party's perspective
- Prominent legal disclaimer built into every report, with honest positioning as informational rather than legal advice
- Pure prompt-based skill with no external script dependencies (PDF generation belongs to the sibling legal-report-pdf skill)
- AI output is not legal advice and cannot substitute for a licensed attorney; jurisdiction-specific variation is not guaranteed
- No license information in the repository (License: unknown), creating uncertainty for commercial use
- No test suite and no documented user validation of analysis quality
- Relies on Claude Code's slash-command trigger and Read/WebFetch tool references; minor edits needed to port elsewhere
- README efficiency and cost claims (60-second review, $300-$500/hour saved) are marketing figures, not verifiable facts
How do you install this skill?
Install the whole 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. The repo does not document a manual install for the legal-risks skill alone; manually, you would copy the skills/legal-risks/ folder into your Claude Code skills directory (the exact directory path is not specified in the source). License is unknown.
How do you use this skill?
In Claude Code, run /legal risks <file>, where <file> can be: a file path (read via the Read tool), a URL to a contract document (fetched via WebFetch), or contract text pasted inline. The skill analyzes the contract and generates RISK-ANALYSIS.md in the same directory as the input file, or the current working directory for pasted text.
How does this skill compare with similar options?
Within the same repository, /legal review is the flagship command using 5 parallel agents and producing a 0-100 Contract Safety Score; this skill (/legal risks) is more focused — a single deep pass of clause-by-clause scoring and hidden-risk hunting. No third-party competitors are named in the source.