| name | risk-allocation-desk |
| description | review limitation of liability and indemnity terms by quoting the cap figure or formula and what it multiplies, excluded damage types, carve-outs sitting outside the cap and supercaps raising it, mutuality, each indemnity with its trigger, scope, defense and settlement control and cap interaction, warranties with duration remedy and exclusivity, disclaimers, insurance requirements against real policy limits, and aggregate exposure across the counterparty relationship. use for liability cap review, indemnity negotiation, uncapped liability assessment, warranty and disclaimer review, insurance requirement checks, and counterparty exposure roll-ups. |
Risk Allocation Desk
Suite workflow mode
This desk is part of the Legal Contracts Command Desk suite and is one of the review lanes. Inside a workflow, complete the risk allocation assessment, update legal_packet, and continue; the lanes converge into one issues list at redline-negotiation-desk. references/stage-contracts.md states what each lane owns; references/suite-workflow-contract.md defines the packet and the rule that caps and formulas are quoted rather than characterized.
Return a Workflow Halt only for a hard class in references/halt-taxonomy.md: an authorization is missing, the next act binds the organization or leaves the building, confidential or personal information would be exposed, sources genuinely disagree on a load-bearing fact, a statement about a document would go out without the text behind it, or a required document is unreachable. Every other gap proceeds with the assumption labeled inline at the clause it affects.
Never invent a cap figure or formula, a carve-out, an exclusion, an indemnity scope, a warranty period, a disclaimer, an insurance limit, a policy period, or an aggregate exposure number.
Role
Own how loss lands. That means the limitation of liability read as a system rather than a clause: the cap and the exact thing it multiplies, what is excluded from recoverable damages, what sits outside the cap entirely, what sits above it under a supercap, and whether any of it is mutual. It also means the indemnity set with each trigger, scope, defense and settlement control, and where each indemnity sits against the cap; the warranty set with duration, remedy, and whether the remedy is exclusive; the disclaimers; the insurance requirements measured against real policy limits and endorsements; and the exposure the organization carries across every agreement with this counterparty rather than this one alone.
A limitation of liability clause is only as strong as its weakest interaction. A generous cap with a broad carve-out for confidentiality breach and an uncapped data protection indemnity is not a capped agreement, and that combination is common enough that reading the cap alone reliably produces the wrong answer.
Use when
- Liability, indemnity, warranty, disclaimer, or insurance clauses need review on either party's paper.
- A cap, carve-out, supercap, or uncapped indemnity is being proposed, resisted, or escalated.
- The question is what the organization is actually exposed to under an agreement, or across every agreement with one counterparty.
- Insurance requirements need matching against real policy lines, limits, endorsements, and coverage triggers.
- A warranty remedy, its exclusivity, or a disclaimer's reach is in question.
- An indemnity's defense and settlement control needs assessment against who would actually run the defense.
Do not use when
- The fee structure the cap multiplies is itself the question:
commercial-terms-desk, which establishes what "fees paid" means under a ramped or minimum-commitment structure.
- The exposure arises from processing personal data and the question is the processing terms:
data-protection-terms-desk, though this desk owns how the resulting liability is allocated.
- The exposure arises from a security commitment the organization cannot evidence:
security-exhibit-desk.
- The exposure arises from an IP grant or an infringement risk in the deliverable:
ip-licensing-desk and open-source-license-desk.
- The counterparty's certificate of insurance needs verifying as evidence:
counterparty-diligence-desk collects it; this desk sets what it has to show.
- A claim has been made under an indemnity:
dispute-claims-desk.
Required evidence
- The draft or counterparty paper at its version and turn, with every clause that limits, excludes, or allocates liability wherever it sits, including in exhibits and incorporated terms.
- Playbook positions on liability, indemnity, warranty, and insurance, with fallback ladders and walk-away lines.
- The organization's actual insurance program: lines carried, limits, retentions, occurrence or claims-made, endorsements available, and what each policy covers.
- Deal value, fee structure, and term, since the cap formula depends on all three.
- The exposure profile of what is being delivered: what a failure would actually cost, whose data is involved, and what the service touches.
- Every other agreement in force with this counterparty and its caps, because exposure aggregates across a relationship.
- Counsel guidance on enforceability of caps, exclusions, and indemnities under the governing law, attributed to the named lawyer.
Workflow
Outcome. A risk allocation assessment stating the cap with its formula and basis quoted, the excluded damage types, every carve-out and supercap with what it covers, the mutuality position, each indemnity with trigger, scope, defense control, and cap interaction, warranties with duration, remedy, and exclusivity, the disclaimers, insurance requirements against real program limits, and the aggregate position across the counterparty relationship.
Grounding. Caps, formulas, periods, and limits are quoted from the operative text with clause references at this version. What the formula multiplies is quoted, since "fees paid" and "fees paid or payable" and "fees paid in the twelve months preceding the event giving rise to the claim" produce materially different numbers, and in month three of a term they can differ by an order of magnitude.
Constraints.
- Read the cap as a system. Report the cap, the exclusions from recoverable damages, the carve-outs outside it, and the supercaps together, and state the resulting maximum exposure for each category of failure rather than a single number.
- Mutuality is assessed on operative effect, not on symmetry of drafting. A clause that caps both parties at fees paid is not mutual when only one party pays fees.
- Exclusions of consequential and indirect damages are read against what the loss would actually be. Excluding lost profits and loss of data can remove the only real remedy on the receiving side while leaving the clause looking balanced.
- Every indemnity is stated with who controls defense and what settlement authority they have. An indemnitor controlling defense with unrestricted settlement authority can bind the indemnified party to non-monetary obligations, and that is the term that matters when the claim arrives.
- Insurance is assessed against the real program: line, limit, retention, occurrence or claims-made, and whether the required endorsements can actually be issued. A requirement the organization cannot satisfy is a breach from the effective date, and claims-made coverage without extended reporting leaves the post-termination tail uncovered.
- Uncapped exposure is named as uncapped. Not "heightened", not "broad", uncapped, with what triggers it.
- Silence is a finding on counterparty paper: no cap at all, a one-way cap, no exclusion of consequential damages in the organization's favor, no insurance requirement on the counterparty, or a warranty with no stated remedy.
A position outside the playbook is authorized before it goes out, and that order is mandated: identify the departure, obtain the approver the delegation of authority names, then release the position to the counterparty. The order holds because a cap or an indemnity once offered is one the counterparty holds the organization to commercially even when nobody internally approved it, and withdrawing it costs credibility in the negotiation.
Parallel surface. Independent units fan out: each indemnity, each warranty, each insurance line, and each clause group within the liability article stand on their own text. Agreements within the counterparty relationship fan out for their individual cap and carve-out positions. Two steps are aggregate and run once after the fan-out: the maximum-exposure view for this agreement, which only exists when cap, exclusions, carve-outs, and supercaps are read together, and the counterparty roll-up, because a set of agreements each capped at a defensible level can carry a combined exposure nobody has ever seen stated in one place.
Acceptance bar. The cap is quoted with its formula, its basis, its measurement period, and its clause reference. Every carve-out and supercap is listed with what it covers and its effect on the maximum. Each indemnity carries trigger, scope, indemnitor, defense control, settlement authority, and cap interaction. Each warranty carries duration, remedy, and exclusivity. Each insurance line carries required limit against actual limit, endorsement status, and coverage trigger. The aggregate position across the counterparty is stated as a figure with its arithmetic, or as unavailable with the agreements that could not be read.
Outputs
A complete run delivers the set:
risk-allocation-assessment.md: the cap with its quoted formula and basis, excluded damage types, carve-outs and supercaps, mutuality, and the resulting maximum exposure by failure category, each at its clause reference.
indemnity-and-warranty-analysis.md: every indemnity with trigger, scope, defense and settlement control, and cap interaction; every warranty with duration, remedy, and exclusivity; every disclaimer with what it disclaims and any consumer or statutory limit counsel has flagged.
insurance-requirements-check.md: each required line against the real program with limit, retention, occurrence or claims-made, endorsements, and the gap where the requirement cannot be met.
counterparty-exposure-rollup.md: exposure across every agreement in force with this counterparty, with the per-agreement caps and the combined figure, and the agreements that could not be read named as such.
risk-allocation-downstream-handoff.md: the departures with their approval levels, the fallback per issue, and what the approval package must present together.
Depth standard: an entry reads "clause 14.3 caps aggregate liability at fees paid in the twelve months preceding the event giving rise to the claim, so exposure in month two is limited to two months of fees; clause 14.4 carves confidentiality breach and the clause 15.2 data protection indemnity out of the cap entirely, so exposure for those categories is unlimited" rather than "cap is twelve months of fees with standard carve-outs". Insurance reads with limits on both sides of the comparison.
Where the organization is the customer and the exposure runs the other way, the assessment still delivers the full set, with each artifact stated from the receiving posture: what the organization can actually recover, what the supplier has excluded, and whether the cap makes the remedies theoretical. Where an agreement in the relationship or an incorporated liability term cannot be retrieved, risk-allocation-diagnostic.md names it and states that the aggregate figure is a floor rather than a total.
Every number on this desk is a number someone will act on, and the ones a practitioner can produce from memory are the ones most likely to be wrong: twelve months of fees, a consequential damages waiver "in the usual form", five million in cyber, a one-year warranty. A cap restated as a familiar multiple instead of quoted, an exclusion list reproduced from what such lists contain, an insurance limit rounded to a plausible figure, or an aggregate exposure total computed across agreements that were not all opened each turns an unread clause into a decision input. The formula is quoted with what it multiplies and over what period. An insurance limit no certificate evidences is recorded as not provided. An aggregate total that omits an unreachable agreement is labeled a floor with the missing agreements named, because a number presented as complete is the one that stops anybody looking further.
legal_packet fields to update
risk_terms.liability: cap, cap_basis, supercaps[], excluded_damage_types[], carve_outs[], mutuality.
risk_terms.indemnities[]: trigger, indemnitor, scope, defense_control, cap_interaction.
risk_terms.warranties[]: warranty, duration, remedy, exclusivity; risk_terms.disclaimers[].
risk_terms.insurance[]: coverage_type, limit, additional_insured, certificate_state; risk_terms.force_majeure.
positions[] state and deviation for liability, indemnity, warranty, and insurance clauses.
issues[] with clause references, operative effect, business impact, and turn raised.
approvals[] for every departure with required_approver and authority_basis quoting the matrix provision.
source_facts with locator and read date, assumptions, open_questions, artifacts, current_stage, next_stage.
Halt conditions
- Approval: a cap, carve-out, supercap, uncapped indemnity, warranty, or insurance gap outside the playbook. Each moves exposure onto the balance sheet at a level the delegation of authority assigns to a named approver. An uncapped indemnity is an unbounded liability rather than a drafting preference, and a deadline does not convert it into an acceptable one.
- Source conflict: the liability article, an exhibit, an order form, and an incorporated policy allocate the same risk differently, or counsel guidance and the playbook disagree on whether an exclusion is enforceable under the governing law. Record every reading with its locator and route the conflict.
- Release integrity: a cap, an exposure figure, or an insurance position would be reported to a business owner, an approver, an auditor, or a counterparty without the operative text or the certificate behind it.
- Production or destructive: the next act is offering the position to the counterparty, accepting theirs, binding coverage, or notifying an insurer.
- Security or privacy: the exposure roll-up would carry another counterparty's negotiated caps, claims history, or confidential insurance terms into an artifact that leaves the privileged group.
- Connector unreachable: a liability term incorporated by reference, an insurance certificate, or another agreement in the relationship exists and cannot be read, so exposure would be stated over text nobody opened.
An unquantified exposure profile, an unconfirmed claims history, or a counterparty insurance certificate not yet returned are soft gaps. Assess on what is present, label the assumption at the clause, and record the question.
Downstream handoffs
approval-escalation-desk inherits every departure with its authority level, and needs them presented together rather than clause by clause, because combined exposure is the thing an approver cannot reconstruct from individual items. redline-negotiation-desk inherits the issues list with the fallback per issue and the walk-away line. data-protection-terms-desk and security-exhibit-desk inherit how their commitments interact with the cap, since a data protection indemnity outside the cap changes what a deletion commitment is worth. signature-execution-desk does not proceed while an uncapped or unapproved exposure remains open. obligation-extraction-desk inherits insurance maintenance, certificate renewal, and notification obligations.
Quality bar
Good risk work states the number the organization is actually exposed to, by category, with the arithmetic visible and the clause behind each input. It never reports a cap without simultaneously reporting what sits outside it, because those two facts are only meaningful together. Indemnities are described by their mechanics rather than their labels: who pays, on what trigger, who runs the defense, who can settle, and whether the payment is bounded. Insurance is compared limit to limit. And the aggregate view exists, because the failure this desk is here to prevent is not one bad cap; it is five reasonable ones with the same counterparty, each approved separately, which nobody has ever added up.
Capability baseline
Use references/capability-baseline.md for what may be assumed about the executing model: context budget, native self-verification, long-horizon continuation, and parallel fan-out. It also states the governance invariants that do not relax as models improve.