Combine workflows, customize for your firm, and quality-check AI output.
5 min readMost legal matters do not fit neatly into a single workflow. An M&A transaction involves NDAs, corporate documents, employment agreements, and data protection assessments. A SaaS deal requires contract review followed by a data processing agreement check. Skills are designed to be chained.
Sequential chaining. Run one Skill, review the output, then apply the next. For an acquisition target, you might start with the NDA Review skill on the confidentiality agreement, move to the GmbH Formation skill to verify the corporate structure, and finish with the Employment Reference skill on key personnel files. Tell your AI agent when you are switching: “Now apply the GmbH Formation skill to this Gesellschaftsvertrag.”
Complementary pairing. Some Skills naturally follow each other. After running a SaaS Contract Review, apply the Data Processing Agreement skill to the accompanying DPA --- the first Skill may flag data protection concerns that the second Skill can investigate in depth. After an Employment Contract Review, use the Employment Reference skill to cross-check whether the reference aligns with the contractual terms.
Building a matter workflow. For recurring matter types, document the Skill sequence your team should follow. A standard technology transaction might always run in this order: (1) NDA Review, (2) SaaS Terms Review, (3) DPA Review, (4) Compliance Check. Writing this down turns individual Skills into a repeatable firm process.
Library Skills are starting points. Every firm has its own standards, preferred language, and client-specific requirements. Customization is where Skills become truly powerful.
Fork and adapt. Download a Skill from the library and create your firm’s version. Add steps that reflect your internal standards --- for example, if your firm always checks for a specific liability cap threshold in vendor contracts, add that as an explicit step.
Embed client preferences. If a key client requires specific clause language or has standing instructions about risk tolerance, build those into a client-specific Skill variant. An NDA Review skill for Client A might include: “Flag any confidentiality term shorter than five years --- Client A requires a minimum five-year term per their procurement policy.”
Add internal references. Point steps to your firm’s precedent documents, clause libraries, or internal guidance notes. A step might read: “Compare the limitation of liability clause against our standard position (see firm template NDA v4.2, clause 8).” The AI agent will not access your internal systems, but the lawyer running the Skill will know exactly where to look.
Version control your customizations. Track changes to firm Skills in Git or another versioning system. When a regulatory change requires an update --- a new court ruling on non-compete enforceability, a GDPR enforcement decision that changes best practice --- you can update the Skill once and every future use reflects the change. The version history shows exactly what changed and when.
AI-assisted analysis is fast and structured. It is not infallible. Every Skill output requires professional review before it informs client advice.
Verify statute citations. AI models can cite the wrong section number, reference a repealed provision, or conflate similar statutes from different jurisdictions. When the output references § 307 BGB, confirm that the provision is correctly cited and that the analysis matches the current legal position. This takes seconds and prevents embarrassing errors.
Cross-check against current case law. Skills reference statutes and established principles, but the law moves. A Skill written in January may not reflect a BGH decision from March that changes the interpretation of a key provision. After running a Skill, ask yourself: has anything changed since this Skill was last updated? Check the Skill’s version date in the frontmatter.
Watch for hallucinated specifics. AI models sometimes generate plausible but fabricated details --- a court decision that does not exist, a statutory threshold that is slightly off, a deadline that applies in a different jurisdiction. The more specific the claim, the more important it is to verify.
AI structures, you decide. The Skill ensures nothing gets skipped. The AI produces a thorough first-pass analysis. But the final assessment --- the risk evaluation, the strategic recommendation, the client advice --- is yours. Treat the output as a well-organised associate’s memo, not as a finished product.
Skill output is a powerful internal tool. It is not client-ready in raw form.
Use output as a starting point for memos. The structured analysis gives you a solid foundation. Reorganise it into your firm’s memo format, add your professional assessment, and tailor the language to the client’s level of sophistication. A general counsel at a DAX company expects different framing than a founder setting up their first GmbH.
Never send raw AI output to clients. Even when the analysis is accurate, unedited AI output lacks the professional judgment, contextual awareness, and tone that clients expect from their lawyers. It may also contain hedging language or structural artefacts that undermine credibility.
Leverage Skills for client education. The step-by-step structure of a Skill output can help explain your analysis to clients. Walking a client through “Here is what we checked and what we found at each stage” builds confidence in the thoroughness of your review.
As your team adopts Skills, organisation becomes important.
Organise by practice area. Create a folder structure that mirrors your firm’s departments: contract law, employment law, corporate, data protection, compliance. Each practice group maintains the Skills relevant to their work.
Assign Skill owners. Every Skill should have an owner --- a senior lawyer responsible for keeping it current. When the law changes, the owner updates the Skill. When a team member identifies an edge case the Skill does not cover, the owner decides whether to add it.
Version and review on a schedule. Set a quarterly review cycle. The Skill owner checks whether recent legislative changes, court decisions, or regulatory guidance require updates. Document the review even if no changes are made --- this creates an audit trail showing the Skill was assessed as current.
Share across offices. If your firm operates in multiple locations, Skills provide a way to standardise methodology while allowing local adaptation. A base NDA Review skill can have jurisdiction-specific variants for Germany, Austria, and Switzerland.
Track adoption. Monitor which Skills your team actually uses. A Skill that no one runs either needs better training, covers a workflow the team does not encounter, or needs improvement. Usage data helps you invest in the Skills that deliver the most value.
Skills excel at structured, repeatable analysis. Some matters require a different approach.
Complex multi-jurisdictional matters. A Skill tagged for German law will not account for the interaction between German, EU, and third-country regulations in a cross-border data transfer. For matters spanning multiple legal systems, use Skills as a starting point for each jurisdiction but rely on specialist lawyers to assess the interactions.
Novel legal questions. When there is no established methodology --- a first-of-its-kind AI liability dispute, an untested regulatory framework, a novel corporate structure --- a Skill cannot guide the analysis because the analysis has not been done before. Skills encode existing expertise. They do not generate new legal theory.
High-stakes litigation. Trial preparation, witness examination strategy, and settlement negotiations involve judgment calls that depend on facts, personalities, and procedural context no Skill can capture. Skills can help with discrete sub-tasks --- reviewing a contract at issue, checking a statutory deadline --- but the litigation strategy remains entirely human.
Matters requiring creative structuring. Tax planning, complex joint ventures, and bespoke financing arrangements require the kind of lateral thinking that comes from experience and creativity. A Skill can verify that the chosen structure complies with specific requirements, but it will not invent the structure in the first place.
The goal is not to automate legal practice. It is to automate the parts of legal practice that benefit from structure and consistency, so you can spend your time on the work that demands expertise, judgment, and creativity.
We're building a hub where legal experts can build and distribute their workflows. Be among the first contributors.
Submit Your Skill →