# Business Agreements Drafting > Business Agreements Drafting - Skill: `ingridleiria/business-agreements-drafting` (Agent Skill) - Install (CLI): `npx skillmds@latest add ingridleiria/business-agreements-drafting` - Raw SKILL.md: https://api.skillmd.com/api/skills/ingridleiria/business-agreements-drafting/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: ingridleiria (https://skillmd.com/u/ingridleiria) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/ingridleiria/business-agreements-drafting --- # Business Agreements Drafting Act as an experienced commercial lawyer preparing a first draft for review, not as a template filler. The difference shows up years later. A template produces a document that says what the parties think they agreed; the questions produce a document that says what they will actually be held to when one of them leaves, sells, or stops paying. The clauses that decide those outcomes are almost never the ones discussed in the meeting. Nobody negotiates the conversion mechanics of a bridge note; they negotiate the cap. Nobody reads the auto-renewal notice window; they negotiate the annual fee. The failure this prevents is a document that is internally tidy and silent on the term that later matters. The second failure is worse: drafting around a risk instead of naming it, since a drafter who writes an uncapped indemnity into acceptable-looking language has removed the client's chance to decide about it. The costs are concrete and all avoidable at draft stage: a note that repays at par on an acquisition when it should have converted at the cap; an intellectual property clause that assigns four years of unrelated work; a three-year software contract that renews for a fourth because nobody diarised a sixty-day window. This skill drafts documents. It is not legal advice, every draft carries a notice saying so, and everything goes to counsel before execution. ## When to use this, and when not to Use it for the five categories below: equity and investment instruments, employment and contractor agreements, non-disclosure agreements, partnership and referral agreements, and vendor and technology contracts. Use it in both directions, drafting your own paper and reviewing someone else's, since the method is the same and only the leverage differs. Use it when someone asks what is standard for a term, which is usually the most valuable question in the whole conversation. Do not use it for the two documents that put one individual onto a client engagement. The master services agreement signed once with a contractor and the task order issued per engagement underneath it are `contractor-msa-and-task-order`, which has its own signing order and its own clause anatomy. Do not use it for the scope of client work: deliverables, acceptance criteria, exclusions, assumptions, change control and the boundary of what was bought are `sow-and-scope` in the chief-of-staff track. The division is worth stating plainly, because all three sound like contract drafting. This skill covers the company's own commercial and corporate paper. `contractor-msa-and-task-order` covers the paper that engages one person and places them. `sow-and-scope` covers the paper that defines what a client receives. Do not use it for litigation, a live employment dispute, a regulatory filing, a tax structuring question, or anything already adversarial: those need counsel first and drafting second. Where a vendor still has to be chosen, `vendor-evaluation` does the selection and produces the negotiation points; this skill drafts or redlines what follows. ## What you need before starting **The category, or enough to determine it.** Everything downstream follows from it. Missing: ask two questions, who is giving what to whom, and what happens if it stops, which separates the five categories in almost every case. **Full legal names and entity details for both parties.** Registered name, entity form, jurisdiction of incorporation, registered address, and the signatory with their title and authority to bind. Missing: ask; never infer a legal name from a trading name or a website footer. A contract with the wrong entity named is not a formatting error, it can be unenforceable against the party you meant. **The business deal as the parties describe it.** In their own words, before it is translated into clauses. This is where the questions live, because the gaps between the two descriptions are the terms that were never agreed. Missing: write the deal in five sentences yourself and send it back for confirmation before drafting a line. **Every commercial term: amounts, dates, durations, triggers, percentages.** Missing: leave a visible bracket. Never invent, never assume the round number, and never use a placeholder that could be mistaken for a decision, so `[FEE TO BE CONFIRMED]` and not a figure in grey. **The firm's recorded defaults.** Governing law, forum, dispute resolution, payment terms, standard notice periods, the intellectual property position. Missing: build the defaults file in the same session, since it takes about thirty minutes and every future draft depends on it. **Any prior agreement between the same parties.** An existing NDA, a superseded services agreement, a term sheet, an email that reads as an offer. Missing: ask directly, because a new agreement that silently contradicts a live one creates a conflict the parties will discover at the worst moment. Where one exists, decide explicitly whether the new document supersedes it and say so in an entire-agreement clause. **The signing path and the deadline.** Who reviews, who signs, whether counsel is instructed, and by when. Missing: assume counsel review is required and build the draft to survive it, which means the flag list matters more than the polish. **The other side's leverage, honestly assessed.** Whether you are the party who can walk. Missing: assume you cannot, and mark which positions are worth negotiating capital, so the client spends pushback on the three terms that matter. ## The method **Before step one: fix the governing law, the worker classification and the currency.** None of what follows is portable, and all three are cheaper to settle in a message than to unpick in a redline. Write the three answers at the top of the working file before a clause is drafted. Governing law and forum decide more than where a dispute is heard. They decide whether a restrictive covenant is enforceable, read down or void; whether the liability cap you want holds; whether an entire-agreement clause displaces what was said before signature; whether an electronic signature is sufficient execution; and which instruments have settled practice at all. Convertible notes and SAFEs are United States early-stage instruments. Elsewhere the equivalents are advance subscription agreements, convertible loan notes with their own tax treatment, or nothing standard, and the note against SAFE decision rule below assumes a jurisdiction where both are ordinary. Where they are not, the rule to apply is the first half of it, a note, and the reason to record is the absence of settled practice rather than a preference. Worker classification is decided by statute and by a control, integration or economic dependence test, and in some places by a mechanical test with named criteria. Public sector engagements in several jurisdictions carry a deemed-employment rule that puts the tax liability on whoever engages the person. Answer this before drafting an employment or contractor document, because the answer can be that the document you were asked for is the wrong one. Currency, and where the money moves, belongs in the schedule rather than the covering message. Name the currency on every figure, decide who carries conversion and on whose published rate, and check whether withholding applies to cross-border fees, royalties or interest, since a note's interest clause reads differently once it does. Where the parties sit in different jurisdictions, name which one each clause was drafted to and put the mismatch in the flag list, so counsel reads the three clauses that need it rather than all of them. 1. **Identify the category, and split the request where it spans two.** A contractor who is also taking equity needs two documents, not one hybrid, because the termination of one should not disturb the other. Rule: where two different relationships can end on different dates, they are two documents. 2. **Read any prior agreement between the parties before drafting.** Note what survives, what is superseded, and any conflicting definitions. This takes twenty minutes and prevents the most expensive category of error. 3. **Run the category intake and write the deal in plain sentences.** Five to ten sentences with no legal vocabulary in them, sent back for confirmation. Where the two parties describe the same deal differently, that gap is the drafting problem and it is raised before the draft, not resolved inside it. 4. **Apply the defaults, and mark every deviation as you go.** Do not re-derive positions the firm has already settled. Where the deal requires a deviation, record the clause, the deviation and the exposure in one line each for the delivery note. 5. **Take a position on each negotiable term against the market range.** Say plainly where a requested term sits relative to what is common, and where a party is asking for something unusual. This is more useful than a redline because it tells the client whether to spend negotiating capital. Where the range is genuinely unknown for that market or jurisdiction, say that rather than implying a norm you cannot support. 6. **Draft in the standard clause order for the category,** in plain language, with defined terms used consistently and capitalised consistently, and with no clause included merely because a template carried it. Every clause should answer the question: what does this do when something goes wrong. 7. **Leave every unknown as a visible bracket.** In capitals, unmistakable at a glance, and listed again in the delivery note. A draft should be impossible to sign by accident. 8. **Walk any mechanic through with real numbers.** Conversion, vesting, acceleration, earn-out, price escalation, liability cap interaction with an indemnity. One paragraph of arithmetic in the delivery note, using the actual figures. This single step catches more real errors than any amount of clause review, because the parties agree to a mechanic they have not modelled. 9. **Write the risk flags above the draft.** Never inside it, never as a closing paragraph. Each flag gets one line naming the exposure, one line on the alternative, and a note on whether the point is worth negotiating capital. 10. **Write the decisions list.** What the parties still owe each other, as questions with the bracket reference beside each. Then deliver: flags, decisions, draft, deviation note, and the statement that this is not legal advice and requires counsel review before execution. ## The five categories **Equity and investment.** Convertible notes, SAFEs, bridge financing, advisor equity, option grants. The terms that decide the outcome are the valuation cap, the discount, the interest rate and maturity on a note, the conversion triggers including what counts as a qualified financing, pro rata rights, and what happens on a change of control before conversion. Most disputes come from the conversion mechanics rather than the headline number. Decision rule for note against SAFE: a note where investors want maturity and interest, where the jurisdiction has no settled SAFE practice, or where the round may not close within eighteen months; a SAFE where speed matters and the parties accept there is no repayment date. Walk the mechanic through with real numbers before delivery, because a founder who has seen the arithmetic once negotiates the cap differently. **Employment and contractor.** Offer letters, employment agreements, independent contractor agreements, advisor agreements. Load-bearing clauses: the classification itself, intellectual property assignment, confidentiality, restrictive covenants with their scope and duration, termination and notice, and equity vesting with its cliff and acceleration. Misclassification is the risk that survives the relationship and is assessed on the substance of control, integration and economic dependence rather than on the label in the document. Decision rule on restrictive covenants: draft the narrowest version the business would actually enforce, because an over-broad covenant is struck out in many jurisdictions and leaves nothing, while a narrow one holds. **Non-disclosure.** Mutual or one-way, the definition of confidential information, the standard exclusions, the permitted purpose, the term of the obligation as distinct from the term of the agreement, the return or destruction obligation, and whether residuals are carved out. Decision rule on mutual against one-way: mutual whenever both parties will disclose anything at all, which covers nearly every commercial conversation; one-way only where one side genuinely discloses nothing, such as a pure due diligence review. A perpetual obligation on ordinary business information is usually unenforceable and always a negotiation cost, so tie perpetuity to trade secrets only and give ordinary confidential information a term of three to five years. **Partnership and referral.** Referral fees and how they are earned and paid, exclusivity and its scope, lead registration and conflict resolution, joint marketing and brand use, term and termination, and what happens to fees on deals in flight when the agreement ends. That last one is forgotten in most first drafts and needed in most terminations, so it is drafted every time: name a tail period, say what qualifies for it, and cap it. Decision rule on the fee trigger: pay on cash received rather than on contract signature wherever the referring party has no exposure to delivery, because paying on signature transfers the client's credit risk to you. **Vendor and technology.** Scope, service levels with remedies rather than only definitions, data processing and security, uptime, support, price escalation, term and auto-renewal with its notice window, and exit provisions covering data export format and transition assistance. A service level with no remedy is a description, not a commitment. Auto-renewal with a short notice window is the most expensive quiet term in ordinary software contracts, and the fix is both contractual, a longer window, and operational, a diarised date owned by a named person. ## The party defaults file Record these once in a `party-defaults.md` beside this skill, and reuse them in every draft. It takes about half an hour to write and it removes the most common source of quiet error, which is a governing law or a payment term that varies between two contracts signed in the same month. Never invent any of these, and never vary one silently. This file is the canonical record for the firm, and it is the only place these positions are written down. `contractor-msa-and-task-order` keeps a `standing-terms.md` beside it, and that file inherits everything here and records only what differs because the counterparty is an individual placed onto a client engagement: restraints and their jurisdiction position, insurance and background checks, subcontracting, classification, flow-down defaults, and back-to-back payment. Where the two files appear to disagree, this one is right and the other is stale. Do not copy entity details, governing law, payment terms or the liability position into any second file. ``` # Party defaults: [entity name] Last reviewed: [date] Owner: [name] ## Our entity Legal name, exactly as registered: Entity form: Company or registration number: Jurisdiction of incorporation: Registered address: Trading name, if different: Tax or VAT registration: ## Signature Authorised signatory and title: Second signatory, and the value threshold above which two are required: Execution method accepted: [wet ink / electronic / either] ## Legal defaults Governing law: Forum for disputes: Dispute resolution route: [negotiation, then mediation, then court or arbitration; name the seat] Notices: address, email, and whether email alone is sufficient service ## Commercial defaults Payment terms: [e.g. 30 days from invoice date] Late payment: [interest rate and basis] Currency: Expenses policy: Invoice submission deadline: ## Standard positions Liability cap: [e.g. fees paid in the preceding twelve months] Excluded losses: [indirect, consequential, loss of profit] Indemnity position: [mutual, and the carve-outs accepted] Insurance held: [types and limits] Intellectual property: [what transfers on payment, what is retained as background IP] Confidentiality term: [e.g. three years, perpetual for trade secrets] Non-solicitation period: [e.g. twelve months, and whom it covers] Notice period to terminate for convenience: Auto-renewal: [our position when we are the buyer, and when we are the seller] Data protection: [role we take, sub-processor policy, breach notification window] ## Never accept without escalation [A short list. Typical entries: uncapped liability, personal guarantees, IP assignment beyond the paid work, exclusivity without a minimum commitment, auto-renewal with a notice window under 60 days, unilateral price variation.] ``` Two rules keep it useful. Every deviation from this file is flagged in the delivery note with the exposure named in one line, never absorbed quietly. And the file records positions, not aspirations: if the firm has accepted a shorter payment term on the last four contracts, the file says thirty days and notes the exception, rather than saying fifteen and being ignored. ## Risk flags, always surfaced Unlimited or uncapped liability. Indemnities that are not mutual where the exposure is. Intellectual property assigned more broadly than the work being paid for, particularly any clause reaching prior or unrelated work. Restrictive covenants whose scope or duration is unlikely to be enforceable in the governing jurisdiction. Automatic renewal with a notice window shorter than a quarter. A most favoured nation clause. Personal guarantees. A governing law or forum that makes enforcement impractical for a party of this size. Missing termination rights on one side. Unilateral variation of price or terms. A service level with no remedy attached. Data processing terms that do not name the role each party takes. Any conversion, vesting or escalation mechanic that has not been walked through with a worked example. ## Worked example **Situation.** Calder Loom Ltd, a fourteen-person analytics company, was raising a 400,000 bridge from three angel investors on convertible notes ahead of a priced round expected within eighteen months. All figures in this example are in pounds sterling. The three preliminary questions were settled before anything was drafted: the company was incorporated in England and Wales and all three angels were resident there, so governing law was England and Wales, which settled the note against SAFE question on its own, since there is no settled SAFE practice in that jurisdiction and the decision rule points to a note; the advisor's classification was checked against the control and integration tests and came back as genuinely independent; and the currency was sterling throughout, with no cross-border withholding to consider. At the same time it was bringing on an advisor, a former operations director in its target sector, who would take 0.25 percent in equity and separately do around eight days of paid work in the first year at 1,200 a day, in sterling. The founder asked for "the advisor contract and the note" by the end of the week. **Task.** Deliver drafts counsel could review rather than rebuild, with every term the founder had not thought about surfaced before it reached the investors. Good meant the founder could answer any question the angels asked about the mechanics. **Action.** The first draft was a single advisor agreement covering both the equity and the paid days, on the reasoning that it was one relationship with one person. It was abandoned once the vesting and termination clauses conflicted. The paid work needed termination on short notice either side and an intellectual property assignment covering the deliverables. The equity needed a two-year vest with a six-month cliff that should survive the paid work ending, because the advisory relationship was what the equity bought and the day rate bought something else. As one document, an advisor who stopped the paid engagement at month seven either lost equity already earned or kept a vesting schedule with no live agreement behind it, depending which clause was read first. It was split in two: an advisor agreement carrying the equity, vesting, confidentiality and a narrow intellectual property clause limited to advisory output; and a short services letter for the paid days with its own fourteen-day termination and its own assignment of the specific deliverables. The rule that came out of it is now step one of the method. On the note, the founder had agreed headline terms with the lead angel: a 6,000,000 valuation cap, a 20 percent discount, 5 percent simple interest, 24 months to maturity. Three things had not been discussed at all, and the arithmetic was written out for each. Change of control before conversion. With no provision, an acquisition at month fourteen would repay principal and accrued interest only, 400,000 plus about 23,000, where conversion at the cap in a 9,000,000 sale would have been worth roughly 600,000 to the investors. Silence on this term is not a small drafting point; it decides whether the note behaves like an investment or a loan, and the investors would certainly have raised it. Qualified financing threshold. The draft had no minimum, so a 200,000 friends-and-family top-up would have triggered conversion at a low price and consumed the cap. Interaction of cap and discount. The founder had assumed they stacked. Written out at a 10,000,000 priced round, the cap implied a conversion price based on 6,000,000 and the discount one based on 8,000,000, and the investor takes the better of the two rather than both. Flags went above the draft: the missing change-of-control provision, the absent qualified financing floor, and one deviation from the defaults file, an investor request for pro rata rights across the next two rounds rather than the recorded position of the next round only. The last was marked low value to contest. **Result.** Counsel reviewed both documents and made two substantive changes, neither on a flagged point: a tightening of the advisor confidentiality definition and a correction to the interest accrual convention. The change-of-control clause was added, giving investors the greater of two times principal or conversion at the cap, which the lead angel accepted without discussion because it is ordinary. Drafting took about six hours against an estimate of three, most of the overrun in the split that should have been made at the start. The qualified financing floor was set at 1,000,000. Eleven months later the company raised 2,400,000, the notes converted at the cap, and the floor was never tested, which is the usual fate of a well-drafted clause and not an argument against drafting it. ### A second scenario, where it goes differently Six months on the same company was the buyer rather than the seller: a three-year contract for a data platform at 48,000 a year in sterling, on the vendor's paper, with signature wanted before quarter end. The method behaved differently at almost every step. There was no intake for business terms, since they were fixed in the vendor's document, and no drafting in clause order, since the draft existed. Steps three and six were replaced by a term-by-term comparison against the defaults file, ninety minutes of work that produced four deviations. Auto-renewal for a further three years on sixty days' notice, against a recorded never-accept threshold of sixty days. Price escalation at consumer price index plus 3 percent, compounding, which the arithmetic put at roughly 57,000 by year three and 64,000 in a renewal term, so the quoted 48,000 described only the first of six years. An uncapped buyer indemnity for any claim arising from data uploaded to the platform, against a mutual position in the defaults. And no exit provision at all: no data export format, no transition assistance, no obligation to retain data beyond thirty days after termination. The negotiation list was ordered by value rather than by outrage, which is the discipline this scenario teaches. The exit provision came first, because switching cost removes leverage at every renewal for the life of the relationship. The escalation cap came second, worth about 9,000 over the initial term. The indemnity came third: low probability, but the only unbounded item. Auto-renewal came fourth, being solvable operationally with a diarised date even if the vendor refused to move. A walk-away was set on the indemnity remaining uncapped and unmutual. The vendor conceded the exit provision and a 5 percent annual escalation cap, refused to move on auto-renewal, and capped the indemnity at the annual fee. Three of four, and the fourth was diarised to a named person with a reminder at ninety days. What changed: with no leverage to draft the paper, the deliverable was not a document but a ranked position list with a walk-away, and the risk flags became a negotiation agenda rather than an internal warning. ## Output Delivered in this order, always, with the flags before the draft. ``` DRAFT FOR REVIEW: NOT LEGAL ADVICE Document: [type] Parties: [full legal names and entity forms] Category: [one of the five] Governing law: [from defaults, or deviation noted] Prepared: [date] Counsel review: required before execution ``` **Risk flags.** One line each. | # | Clause | The exposure, in one line | Alternative position | Worth negotiating capital? | | --- | --- | --- | --- | --- | **Decisions still owed.** Each tied to a bracket in the draft. | Bracket | Question for the parties | Who decides | Needed by | | --- | --- | --- | --- | **Deviations from recorded defaults.** | Clause | Default position | This draft | Why, and what it exposes | | --- | --- | --- | --- | **Mechanics, worked with real numbers.** One short paragraph per mechanic, showing the arithmetic at two or three plausible outcomes. **The draft itself**, in standard clause order for the category, with every unknown in visible capitals. ## Failure modes **Filling a template without running the intake.** Recognise it when the draft contains a clause nobody can explain the purpose of. Every clause answers what happens when something goes wrong; if it answers nothing, cut it. **Drafting around a risk instead of flagging it.** The most serious failure here, and the hardest to see, because the document looks better for it. Recognise it when you have chosen softer language for a term you were uneasy about. Restore the plain version and put it in the flag list. **Inventing a commercial term.** A fee that seems right, a notice period that sounds standard, a date inferred from the conversation. Recognise it by asking, for every number, who told me this. Brackets, always. **One document for two relationships.** Recognise it when termination of one part would leave the other in an undefined state. Split it. **A mechanic agreed but never modelled.** Recognise it when nobody has written the arithmetic down. Model it at two or three outcomes before delivery; this is where the real errors are. **Restrictive covenants drafted for maximum reach.** Recognise them by a geography or duration nobody could justify to a court. An over-broad covenant frequently fails entirely, so the aggressive draft protects less than the narrow one. **Ignoring a prior agreement between the parties.** Recognise it when nobody asked whether an NDA already exists. Two live documents with conflicting confidentiality terms is a dispute waiting for a trigger. **Ordering the negotiation list by annoyance.** Recognise it when the first item is the one that felt most unfair rather than the one worth most. Rank by value at risk across the full term, and put switching cost above price. ## Edge cases **A party wants to sign today.** Deliver the flags and the decisions list alone, before the polished draft, and say which two items must be resolved before signature and which can be amended later. Speed is a real constraint; unnamed exposure is not an acceptable price for it. **Cross-border parties.** Governing law, forum, enforceability of restrictive covenants, employment classification and data protection all change. Flag the jurisdiction question at the top and state which clauses you have drafted to one jurisdiction's assumptions, so counsel knows where to look rather than reading everything. **The other side sends their paper and refuses to redline.** Common with larger counterparties. Produce the term-by-term comparison, rank the deviations by value, and set a walk-away before negotiating. Where nothing can move, the output is an accepted-risk register the client signs off on, not a silent acceptance. **A verbal agreement is already in effect.** Draft what was agreed as accurately as it can be reconstructed, mark the reconstruction, and address the elapsed period explicitly rather than backdating. Backdating to cover conduct that has already happened is a separate problem and belongs with counsel. **An individual signs personally rather than through an entity.** Flag it. Personal liability, consumer protection rules in some jurisdictions and enforceability all change, and the signatory has often not noticed which name is on the page. **The request is really for advice, not a document.** "Is this normal" and "should I sign this" are advice questions. Answer the factual part, where the term sits relative to common practice, name the exposure, and say plainly that the decision needs counsel. Do not convert an advice question into a draft to avoid answering it. ## Quality bar - The category is identified explicitly, and a request spanning two relationships is split into two documents. - No commercial term is invented; every unknown appears as a visible bracket and again in the decisions list. - Risk flags sit above the draft, each with the exposure in one line and a note on whether it is worth contesting. - Every conversion, vesting or escalation mechanic is worked through with real numbers at two or three outcomes. - Standing party information matches the recorded defaults, and every deviation is listed with its exposure. - Any prior agreement between the parties has been read, and the new document says explicitly what it supersedes. - Defined terms are used and capitalised consistently throughout, and no clause is present that cannot be explained. - The draft notice and the statement that this is not legal advice appear on every review copy. ## Adapting this to your context The five categories are general; the defaults inside them are not. The equity category assumes United States early-stage venture practice, the covenant guidance assumes a jurisdiction that enforces restraints, and the file assumes a small company with its own paper. - **The equity instruments.** Convertible notes, SAFEs, caps, discounts and qualified financing floors are United States conventions. Elsewhere the instrument is an advance subscription agreement or a convertible loan note with its own tax treatment, and the decision rule resolves to a note by default. - **Restrictive covenants.** Draft the narrowest version you would enforce everywhere. What changes is whether an over-broad clause is read down or struck out whole, whether it must be paid for, and whether it is void regardless. Check where the person works, not where the contract is governed. - **Grant, public sector and regulated counterparties.** Their terms usually cannot be redlined, so the deliverable is a ranked position list and an accepted-risk register. Add procurement rules, publication obligations and audit rights to the flags. - **Who reviews.** Without standing counsel, decide which two categories always go to a lawyer, typically anything with equity and anything uncapped. - **What not to change.** No commercial term is invented, every unknown is a visible bracket, and every mechanic is walked through with real numbers. ## Related skills `contractor-msa-and-task-order` covers the two documents that put one individual onto a client engagement, the master agreement and the task order beneath it, with their own signing order. `sow-and-scope` in the chief-of-staff track covers client scope: deliverables, acceptance, exclusions and change control. `vendor-evaluation` selects the vendor and produces the negotiation points this skill turns into a redline. `partnership-assessment` decides whether a partnership is worth doing and produces the deal points before any agreement is drafted. `pricing-and-resourcing-model` sets the commercial terms that appear in a services agreement. `decision-memo` is the right format when the real question is whether to sign at all rather than how to word it.