Saved result
Submitted arguments
{}Native result and verdict
{
"does_not_license": {
"reading": "hammer-coverage-intelligence, vintage 2026-10-01, build 2.\nWhat payers publish that they pay, code by code, and what may be said with it.\n\nA RECEIPT IS REDEEMED AGAINST THAT IDENTITY: name, vintage and build together. Every answer this world hands you carries one, and a receipt naming a different vintage is a different world's. Quote all three beside any figure you take from here.\n\nWHAT TO ASK IT. 25 declared tools, read from this world's own `tools/tools.json`. Each line is the first sentence of the tool's own description; the whole of it, and every refusal it can return, is in that tool's entry in `tools/list`.\n rate_position: What payers publish that they pay for a code, positioned against the public fee schedule that anchors it.\n code is one of: \n payer is one of: aetna | anthem | bcbs-mn | centene | cigna | hcsc | highmark | independence-blue-cross | unitedhealthcare | bcbs-north-carolina | uhc | ibx\n refuses 8 ways rather than guessing, each written out in its `tools/list` entry\n rate_comparison: Two to four codes' payer rates set side by side under one receipt.\n codes is one of: \n payer is one of: aetna | anthem | bcbs-mn | centene | cigna | hcsc | highmark | independence-blue-cross | unitedhealthcare | bcbs-north-carolina | uhc | ibx\n refuses 6 ways rather than guessing, each written out in its `tools/list` entry\n stack_check: Whether two or more codes may be billed together, per the NCCI PTP edit file.\n codes is one of: \n refuses 4 ways rather than guessing, each written out in its `tools/list` entry\n screening_yield: How many people a screening programme would find positive and negative, true and false, and what its positive and negative predictive value would be, given a declared test sensitivity, specificity, prevalence and population size.\n prevalence_basis is one of: administrative | measured | modelled | self_reported\n refuses 5 ways rather than guessing, each written out in its `tools/list` entry\n deferred_versus_avoided: Whether a set of declared cost events are truly eliminated by an intervention or merely moved to a different time, reported as two separate totals that are never added together.\n events.classification is one of: prevented | moved\n refuses 2 ways rather than guessing, each written out in its `tools/list` entry\n interpretation_requirement: Who a payer's own policy says may FURNISH and INTERPRET a code.\n code is one of: \n refuses 5 ways rather than guessing, each written out in its `tools/list` entry\n cohort_figures: How many people, and by what means somebody counted.\n refuses 14 ways rather than guessing, each written out in its `tools/list` entry\n comorbidity_edges: MOVED.\n refuses 1 ways rather than guessing, each written out in its `tools/list` entry\n fee_components: What the Medicare Physician Fee Schedule states for a code: whether the code splits into a professional component and a technical component, and the relative values it states for each.\n code is one of: \n refuses 5 ways rather than guessing, each written out in its `tools/list` entry\n filed_charge: What a hospital filed as its own standard charge for an inpatient admission, under 45 CFR 180.\n code_type is one of: MS-DRG | APR-DRG\n filer is one of: cleveland-clinic | hca-medical-city | hss | mount-sinai | northwell | nyu-langone | vanderbilthealth-com\n severity is one of: 1 | 2 | 3 | 4\n refuses 15 ways rather than guessing, each written out in its `tools/list` entry\n hypertension_evidence: MOVED, and it split in two, because it held two kinds of number.\n refuses 1 ways rather than guessing, each written out in its `tools/list` entry\n prognostic_association: One published observational association per answer: this exposure predicted that outcome, in people nobody randomised.\n exposure is one of: masld_nafld | clinic_bp | nighttime_ambulatory_bp\n outcome is one of: cardiovascular_event | incident_type_2_diabetes | liver_specific_mortality | all_cause_mortality | incident_heart_failure\n stratum is one of: unstratified | higher_fibrosis\n refuses 16 ways rather than guessing, each written out in its `tools/list` entry\n treatment_effect: One published RANDOMISED effect per answer: being assigned this arm changed that outcome's rate in a trial.\n arm is one of: no_prior_cvd | prior_cvd | cardiac_rehabilitation\n outcome is one of: incident_heart_failure | major_cardiovascular_event | all_cause_hospitalisation | all_cause_mortality\n refuses 11 ways rather than guessing, each written out in its `tools/list` entry\n regression_to_mean_expectation: What Galton's arithmetic predicts a cohort selected for being extreme will measure next period, before anybody intervened.\n refuses 4 ways rather than guessing, each written out in its `tools/list` entry\n competing_event_coupling: How much of a readmission rate's movement between two periods is the survivor denominator moving underneath it.\n refuses 4 ways rather than guessing, each written out in its `tools/list` entry\n withhold_earnback: What a measure result is worth under a capitation withhold.\n refuses 4 ways rather than guessing, each written out in its `tools/list` entry\n tournament_position: Where a declared score sits against declared cut points, and how far the FIELD would have to move to cross each one with that score held still.\n refuses 5 ways rather than guessing, each written out in its `tools/list` entry\n measure_reliability: How much of a measured rate is the provider and how much is the sample it was measured on.\n between_provider_variance.from is one of: hospital_variance\n refuses 7 ways rather than guessing, each written out in its `tools/list` entry\n declared_inputs: Bring a customer's own aggregate figures into this world under a declared basis, and get back a handle the calculators read.\n mapping.basis is one of: assumption | cited\n mimetype is one of: application/json | text/csv\n refuses 12 ways rather than guessing, each written out in its `tools/list` entry\n cost_of_care_exposure: Size an exposure over a declared population: what a condition accounts for across a book, over a declared period, on figures the caller declares.\n refuses 3 ways rather than guessing, each written out in its `tools/list` entry\n savings_methodology: Classify a declared savings mechanism as hard or soft against the buyer's own definitions, and say what evidence a case built on it is held to.\n mechanism is one of: unit_cost_reduction | utilisation_reduction | demand_redirection | avoidance_of_future_cost | overpayment_recovery | adherence_improvement\n refuses 2 ways rather than guessing, each written out in its `tools/list` entry\n hospital_variance: How far apart providers actually are on one published quality measure, read off a public measure corpus rather than declared by whoever needed the number.\n measure_id is one of: COMP_HIP_KNEE | Hybrid_HWM | MORT_30_AMI | MORT_30_CABG | MORT_30_COPD | MORT_30_HF | MORT_30_PN | MORT_30_STK | PSI_03 | PSI_04 | PSI_06 | PSI_08 | PSI_09 | PSI_10 | PSI_11 | PSI_12 | PSI_13 | PSI_14 | PSI_15 | PSI_90\n refuses 8 ways rather than guessing, each written out in its `tools/list` entry\n filer_inventory: What one hospital filer carries, so a call that prices something is a call that answers.\n filer is one of: cleveland-clinic | hca-medical-city | hss | mount-sinai | northwell | nyu-langone | vanderbilthealth-com\n refuses 8 ways rather than guessing, each written out in its `tools/list` entry\n candidate_screen: The screen that runs before the pitch rather than after it: the caller's own declared candidate programmes, put through an ordered routing of steps, under one receipt.\n candidates.mechanism is one of: unit_cost_reduction | utilisation_reduction | demand_redirection | avoidance_of_future_cost | overpayment_recovery | adherence_improvement\n sweep.input is one of: between_provider_variance | numerator | denominator\n refuses 6 ways rather than guessing, each written out in its `tools/list` entry\n frontier: Where each part of a goal would be settled, before you spend a call on any of it.\n rests_on.kind is one of: a_price_a_hospital_filed | a_rate_a_payer_published | what_one_filer_carries | a_published_effect_of_an_intervention | how_far_apart_providers_are_on_a_quality_measure | whether_a_difference_is_measurable_at_your_volume | how_many_patients_or_admissions | which_hospitals_are_one_system | what_your_own_contract_pays_you | what_something_costs_you_to_do | your_own_payer_mix_or_volume | whether_a_payer_paid_or_denied | whether_a_payer_covers_a_code | which_provider_is_cheaper_or_better | what_an_avoided_admission_is_worth_to_a_provider | whether_a_device_or_test_works\n refuses 1 ways rather than guessing, each written out in its `tools/list` entry\n\nMOUNTED BY THE RUNTIME, declared by no world and present in every one: judgement_check, claim_check, deliver, session_trace, progress_check, model_advisory, knowledge_navigate, simulate, sensitivity, interval_check, decision_navigate, network_infer, hello_world, say_world.\n\nWHAT IT REFUSES, verbatim from this world's own `cannot_answer`. THESE ARE THE PRODUCT RATHER THAN A LIMITATION TO APOLOGISE FOR: a question on this list has no answer here at any n, by any method, including arithmetic over figures the world does serve. Route it somewhere else or say plainly that it is unanswered.\n 1. a denial rate or a payment probability, from this corpus, by any method, including arithmetic on counts the world does serve; MRF files hold negotiated rates, not adjudications, so a rate's presence does not record a payment and its absence does not record a denial\n 2. whether a payer covers a code; a published negotiated rate says what a payer pays when it pays, not whether it will, so `payer X covers code Y` is not a sentence this corpus can settle\n 3. a cross-payer average without declared weights; the volume-weighted blend in the July rate report was licensed by a declared weighting, an unweighted national average is a number about this corpus's payer mix and not about any market, and the payer mix to weight by is the operator's to declare\n 4. a per-provider payment prediction or a ranking of providers; provider rows are observations of contracted rates, and 30 to 60 percent of small-cell variation is sampling error before anything else is said\n 5. a clinical claim about any device or test; this corpus prices procedures and holds no evidence about whether a product works, and a cited effect size for disease progression under an intervention class is never evidence about a named product's performance\n 6. what was in force as of a date beyond its own vintage; rates refresh monthly, LCDs change weekly, and upstream files are deleted at roughly 90 days, so the vintage a claim relied on has to be named explicitly, and a claim carried across vintages must be rechecked, never inherited\n 7. non-coverage from the absence of a rate row; payers omit, fragment and misfile disclosures market-wide, illustrated by one payer's own disclosure spanning roughly 11 million separate files and another payer's disclosure misusing the schema it is filed under, and the supportable sentence is that no rate was recorded in this month's disclosure, not that the code is not covered\n 8. a Medicare bundling verdict, and no bundling verdict at all for a pair with no edit row; the Medicare NCCI PTP baseline is gated behind an AMA CPT license acceptance form and is not held by this world, so the stacking edits served here are the Medicaid NCCI edition, which CMS publishes ungated, the two editions are not guaranteed to agree, and reporting a verdict from this corpus as `Medicare bundles these` states a rule this world has never read; separately, a pair with no row is UNKNOWN and never permission, since Category III codes routinely carry no row and 0689T against 91200 is exactly that case in the shipped file\n 9. a saving for a named payer's own book; the deferred-versus-avoided arithmetic runs over assumptions and cited transitions the caller declared, and a payer's realised saving depends on their population, their contracts and their utilisation, none of which this world holds, so quoting a modelled figure as that payer's saving states a fact about their business nobody here has measured\n 10. a total that sums a deferred cost with an avoided one; a cost that arrives later is not a cost that never arrives, and any figure adding a timing shift to a prevented event reports a saving that does not exist, which is why the two are returned as separate totals and no combined one is served at all\n 11. who may furnish or interpret a code, from the fee schedule; the PC/TC indicator and the status code say whether a code splits and whether Medicare prices it nationally, they do not say which specialty may furnish the professional component, and reading a restriction out of them would answer the most consequential half of that question with a field that does not contain it\n 12. how many people in a geography have MASLD, nor a prevalence converted between measurement classes; no source this world ships observes a liver below the national level, and administrative and epidemiological prevalence diverge in both directions by amounts that depend on the condition and the algorithm, so every cohort figure names the class it came from and this world holds no factor for converting between them\n 13. a causal effect from a comorbidity edge, a risk for the operator's own population, or a combined risk across more than one edge; every edge served here is observational and adjusted and adjustment is not randomisation, none of the cohorts behind them is a United States commercially insured screening population, and the severity-stratified estimates share their comparator with the unstratified ones so multiplying the two reports a hazard nobody measured\n 14. a causal effect from a hypertension prognostic record; nobody in those cohorts was randomised, adjustment is not randomisation, and a hazard ratio for a pressure predicting an outcome is not the effect of lowering that pressure\n 15. a combined figure across the hypertension corpus's two lists; a prognostic hazard ratio and a randomised risk ratio are different kinds of number measured against different comparators, so no arithmetic over one of each is a quantity anybody measured, and a caller who wants one of each asks twice and reads them side by side\n 16. a Hospital Readmissions Reduction Program penalty, an excess readmission ratio, or any 30-day readmission figure; no record this world holds counts 30-day unplanned readmissions for an index condition, and the nearest record it does hold, all-cause hospitalisation over 6 to 12 months in trial participants, has a different numerator, a different denominator and a different period\n 17. a HEDIS rate, a Star Rating, a quality bonus or a Medicaid capitation withhold; HEDIS measure specifications are NCQA's and licensed, the Star Ratings technical notes and the HRRP methodology are CMS's, none of the three is held here, so this world can report what the clinical literature published and never what any measure counts, which is the half a payment argument actually turns on\n 18. a cardiovascular risk or treatment effect for the operator's own population; every record names the people it was measured in, the cardiac rehabilitation trials recruited under 15 percent women, and the prognostic records are in people already being treated for hypertension\n 19. a hospital's charge compared with a payer's rate; a charge is what a provider filed and a rate is what a payer published, they are different kinds of money, and this world serves them through separate tools so that no single call can ask for one of each and no arithmetic can cross them\n 20. a hospital's filed payer name read as a payer; those names are free text each hospital wrote for itself, 240 distinct strings against the payer slugs the rate tools use, with one company appearing under several spellings, some entries naming a third-party administrator renting a network rather than the insurer, and some naming a government programme, so resolving one to a payer is an inference this world has not licensed\n 21. that a filer does not price an admission because its code is absent from their file; DRG groupers are revised and filers sit on different editions, so a missing code may be a code that edition does not define, and neither the file's schema version nor its vintage names the edition, which makes this the same refusal as declining to infer non-coverage from a missing rate row\n 22. a Cost of Care savings figure, from any money this world holds, by any method, including arithmetic done by hand; a plan's business-case standard scores savings on paid amounts, as opposed to allowed amounts or billed charges, and no compartment here produces a paid amount: payer MRFs hold negotiated rates, what a payer published it pays when it pays, and 45 CFR 180 filings hold filed charges, filed counterparty rates, filed brackets and cash prices, what a hospital published about its own prices, so a savings figure built from any of these and presented as Cost of Care savings is the artefact the buyer's actuarial review exists to reject, and what would settle it is adjudicated claims, which this world does not hold\n 23. that a modelled saving is incremental to a prior period's run rate, expressed on a plan-membership basis, or reported at gross value; the buyer's standard requires all three, and each is a fact about a plan's own book, its run rate, its membership and its accounting, which no corpus here observes, so all three enter a submission as the plan's own declared inputs and never as anything this world verified\n 24. a question from a branch of the medical-cost lever tree this world does not sit on; what payers publish that they pay is contracted-pricing evidence, so this world sits on the reduce-unit-costs branch and only its contracted-pricing leaf, and for every other branch the reason is one of three different facts rather than one word: a pharmacy question meets an absent corpus, since no pharmacy source is held here in any state and nobody has packed the data; an over-payment question meets an absent corpus and a licence wall together, since finding an overpayment needs the adjudications this world does not hold and the bundling-edit half of that work runs into the AMA CPT licence acceptance form the Medicare NCCI baseline sits behind; and a demand question, legitimate or not, meets a question this corpus cannot settle at all, since demand is utilisation, utilisation is adjudicated claims, and a price says nothing about how often anything happens; the branch is named so the question can be taken to a corpus that serves it rather than answered with an adjacent figure that happens to exist here\n 25. what an avoided admission or an avoided readmission is worth to the provider who avoided it; the same avoided admission is revenue that never arrives under fee for service and cost that is never incurred under a risk contract, so the sign of the figure is decided by the contract that provider is paid under, and this world has read no contract: it holds what payers published that they pay and what hospitals published that they charge, and neither records how any provider is paid for any population, so what would settle it is that provider's own contract terms and their own book, which are theirs to declare and never ours to infer from a published price\n\nNAVIGATION ANCHORS THAT RESOLVE IN THIS BUILD, read from the compartment on this world's disk and not from any prose about it. `knowledge_navigate` takes exactly one anchor.\n query RESOLVES a semantic index is present: text search over the indexed chunks answers\n node RESOLVES a numbering is present: a canonical node id from an earlier answer resolves\n select RESOLVES parquet tables attach: a table selector resolves and comes back with the matched rows' columns. `knowledge_navigate`'s own description lists the tables and their columns\n steps PRESENT typed transitions this world's graph carries, each with the way its edge is stored: carried_in (register entry → data compartment), read_by (data compartment → tool), declares (register entry → column name), refuses (tool → refusal), names (refusal → data compartment), gated_by (register entry → access class), addresses (article → procedure or DRG), issued_by (article → contractor), bundled_with (procedure (NCCI column 1) → procedure (column 2, the code denied)), priced_by (procedure → pricing class), rate_observed (payer → procedure or DRG), raises_risk_of (condition → outcome), evidenced_by (effect-size edge → citation), co_occurs_with (condition → condition), covers (register entry → jurisdiction), serves (contractor → jurisdiction), computed_by (tool → estimator), member_of (procedure or DRG → code set), refuses_composition (relation → corpus refusal), asserts (document → claim), about_code (claim → procedure or DRG), co_priced (procedure or DRG → procedure or DRG, stored both ways), manual_names (claim → procedure or DRG), manual_issued_by (manual → program or payer), manual_asserts (manual → claim), filed_charge (hospital → procedure or DRG), files (hospital → payer)\n\nHOW TO READ WHAT COMES BACK. Everything above is routing. These are the rules that decide whether a sentence you write from an answer is licensed.\n A REFUSAL IS A VALUE AND NOT AN ERROR. It comes back as an ordinary successful result carrying `kind: \"refusal\"`, a reason, and a `cannot_be_concluded` naming the sentence the evidence will not support. Retrying it returns the same refusal. Working around it by hand, having read it, is the one failure no check downstream can catch, because a claim you never submit is a claim nothing sees. The refusal is the finding: report it.\n\n `entailed` IS NECESSARY AND NOT SUFFICIENT. It is the only `claim_check` verdict that licenses writing a claim down, and it is not permission to conclude: it says the arithmetic over the figures in that one payload settles that one sentence. A claim resting on a figure whose value your own arguments carried in comes back refused rather than `entailed`, and every such figure is listed in the reply under `world.restates_your_input`. Every other verdict is a refusal with its own remedy, and `claim_check`'s entry in `tools/list` names each one, with what to do about it.\n\n ABSENCE IS NOT ZERO, AND THIS WORLD TELLS ITS ABSENCES APART. It refuses rather than handing back an empty result, because an empty result reads as a zero and a zero is a finding. The kinds it raises, each written out in its own tool's `tools/list` entry:\n it is outside the vocabulary, so nobody ever looked and the absence is unreportable: coverage/code-not-searched, coverage/outcome-not-in-corpus, coverage/stratum-not-in-corpus, coverage/measurement-not-in-corpus, coverage/arm-not-in-corpus, coverage/measure-not-in-corpus\n this build never read it, whatever the source holds: cohort/source-has-no-table, cohort/geography-not-built, cohort/period-not-built\n the publisher held it back: cohort/field-blocked, cohort/every-cell-withheld\n somebody looked and the corpus holds nothing: coverage/edge-not-measured, coverage/record-not-measured, coverage/cited-is-not-measured\n the compartment was never mounted, so nothing could be read at all: coverage/missing-data\n\n BASIS TRAVELS, AND IT ONLY EVER WEAKENS. This world serves a `basis_detail` beside its figures, and the rule it serves there is: A figure computed from inputs of more than one basis carries the weakest basis among them, strongest to weakest being measured, then cited, then assumption, so a figure is never stronger than the weakest input that produced it.\n\n READ A FIGURE AGAINST ITS OWN DISPERSION, which rate_position and rate_comparison report beside every cell. Two figures beside each cell say how thin that cell's own evidence is, and neither is a verdict. stddev_over_median is the cell's own stddev divided by its own median, both of them figures on this same cell: a value well above one says the spread dwarfs the middle, which is a property of the rows behind the cell rather than a finding about what the payer pays. largest_n_over_n is the largest row count any cell at or above the reporting floor carries for this code in this extract, divided by this cell's own row count, and largest_n_for_code names the payer and the count that division used so the arithmetic can be redone from this payload. No threshold is applied to either figure, no cell is withheld or refused on account of either, and this world states no opinion about which values make a cell worth standing behind: that reading is the caller's, on evidence this answer states rather than on prose it does not carry.\n\n\nTHE CLAIM DISCIPLINE. A world-backed answer ENDS IN A CLAIM REPORT: every sentence you wrote from this world, listed beside the verdict it got and the receipt it rests on. An answer without one is an answer whose claims were never put to anything.\nONLY `entailed` LICENSES WRITING A CLAIM DOWN. Every other verdict is a refusal with its own remedy, and neither checker falls back to the other: a checker that quietly fell back to a weaker one would be a fall-through approval with extra steps.\n`claim_check` DECIDES BY ARITHMETIC over the figures in the payload and never reads your sentence, so it is the same claim in any language. `judgement_check` READS THE ENGLISH and decides whether the sentence is one the answer supports. Choose by what you are actually checking, never by which one passed.\nQUOTING A FIGURE EXACTLY IS THE CASE THAT SEPARATES THEM. A decimal in a payload is read as rounded, so `== 1.43` against a payload carrying 1.43 is `not_entailed`: equality would assert a precision the answer never had. That verdict carries the two bounds that WOULD have passed, under `entailed_instead`. To put the published number into a SENTENCE, that is `judgement_check`'s question and it answers there.\nA FIGURE YOU COPY OUT OF ANOTHER ANSWER IS NOT CHECKED BY BEING TYPED IN. `claim_check` decides against one receipt and cannot see where a literal came from, so arithmetic mixing this answer's figure with a number carried over from another is your own arithmetic wearing a verdict. Where a world declares a figure uncombinable, that claim comes back `not_composable` and there is no better number to find: ask for each separately and read them side by side.\nAn answer taken and never checked is the one thing a check on submitted claims cannot see. `deliver` is what puts it in the audit log beside your report instead of nowhere.\n\nTWO WORLDS' ANSWERS DO NOT COMPOSE INTO A THIRD CLAIM. Every answer is redeemed against one world's receipt and THERE IS NO RECEIPT SPANNING TWO, so a sentence built from this world and another is a claim neither world issued and neither checker can check. Report the two findings separately, each against its own receipt, and leave the join to the reader as the reader's own inference.\n\nTHE FULLER SKILLS, where the traps are written out at length. Each is served from THIS build, so it describes these tools and no others: call `hello_world` again with `skill` set to one of the names below and you get that one in full. The description under each name is the skill's own, written to be read on its own, and it is what tells you which to fetch. The URI is what `resources/read` answers from, for a client that would rather hold it as a resource.\n\n skill: hammer-building-the-evidence\n uri: hammer://hammer-coverage-intelligence/skills/hammer-building-the-evidence\n Use when somebody has a commercial goal and needs evidence that survives the buyer's own scrutiny: pricing a service line, pitching a payer, sizing a market, arming a contract negotiation, showing an association's members what their data is worth. Carries the order of operations from a goal to a defensible claim, how to find what data exists before spending a call, the industry's own statistical toolkit mapped to the estimator that serves each, the confounds that have killed real findings in this corpus with what each one cost, how to read a refusal as the next step rather than a wall, and a worked example from a live commercial question to the evidence that answered it. This is the method; the persona skills route the question and hammer-writing-the-deliverable assembles the document.\n\n skill: hammer-cost-of-care-case\n uri: hammer://hammer-coverage-intelligence/skills/hammer-cost-of-care-case\n Use when a plan-side Cost of Care analyst asks whether a use case is worth a slot, what a digital solution saves the plan, how to size an exposure over a declared population, whether a saving is hard or soft, what regression to the mean does to a selected cohort's next period, or anything phrased in PMPM, line of business, funding type or minimum impact threshold. Opens with the refusal that decides these sessions, the paid-amount wall: this world holds no paid amounts and cannot produce a Cost of Care savings figure that would pass the actuarial review a business case goes through. Carries the medical-cost lever tree with this world's position marked on it, the population chain from line of business down to funding type and why fully insured against administrative services only decides who carries the risk, which calculator serves which step of a business case, and the refusals to expect with what each one is telling you. Then the assembly order: name the branch, classify the mechanism before modelling it because hard and soft are different models rather than one model with a label, pin the population, size the savings side as a unit cost times a volume times a rate of change and say which of the three came from where, size the cost side that weak submissions omit (integration, stand-up, member outreach, recurring run cost), and put it on a clock, because a mid-year launch carries a full year of cost against a fraction of a year of savings. Ends on the rates that decide a soft case, which this world holds none of, and the rule that a readmission rate means nothing without its cause rule and time window. Read hammer-savings-classification for the hard and soft definitions in full, and hammer-coverage-navigate for tool-level routing; this is the plan-side framing over both.\n\n skill: hammer-coverage-navigate\n uri: hammer://hammer-coverage-intelligence/skills/hammer-coverage-navigate\n Use when routing a coverage question to the tool that answers it, when you hold a GOAL rather than a question and do not yet know what to ask, or when deciding what this corpus does and does not license you to write. Opens with frontier, the tool that takes a goal in your own words plus the claims it rests on and routes each leaf to one of four places, settled here naming the tool, settled elsewhere naming a corpus in the register, yours to declare because your own book is a fact no corpus holds, or refused permanently together with the nearest question this world does answer. Then a routing table from what an operator is actually asking to the tool that serves it, covering payer rates (rate_position), the same rates for several codes under one receipt so a comparison can be checked (rate_comparison), code pairs against the Medicaid NCCI edit file (stack_check), who a payer's own policy says may furnish and interpret a code (interpretation_requirement), how the Medicare fee schedule splits a code into its professional and technical components (fee_components), what a hospital filed for an inpatient admission under 45 CFR 180, kept apart from payer rates because a charge is not a rate (filed_charge), headcounts and modelled prevalence with their measurement class (cohort_figures), published observational associations, MASLD or blood pressure, one record at a time (prognostic_association), randomised trial effects kept in a separate tool so no call can ask for one of each (treatment_effect), and the calculators that read no data at all (screening_yield, deferred_versus_avoided, regression_to_mean_expectation, competing_event_coupling). Explains the source register that knowledge_navigate reads, why an available source is different from an absent one, and why a register entry can never yield a figure. Names the refused questions with their cannot_answer entries. Covers the traps this corpus invites: percent-of-charge and phantom rows behind the quality filter, professional against technical against global, a rate is not coverage, cross-payer averaging without the operator's own payer mix, the two clocks, and the Medicaid-against-Medicare NCCI edition gap. Explains why positive predictive value collapses at low prevalence and why a deferred cost and an avoided cost are never summed. Every answer closes with the claim report defined in hammer-worlds.\n\n skill: hammer-provider-contracting\n uri: hammer://hammer-coverage-intelligence/skills/hammer-provider-contracting\n Use when somebody asks what a named provider's contracts look like, whether an admission is priced as a DRG case rate or some other way, what a hospital filed for a procedure or an admission, or how a provider's filing compares with what payers publish. This is the working order for that question, one level above the tool surface: which question to settle first and why the obvious order is wrong, the arguments the tools require and the reason each exists, the refusals a real session hits and what each one is telling you, the crossing this world forbids together with the measured case where a careful analyst ran it and got a clean-looking answer that dissolved, how to read a row count in a corpus that replicates, and the checker traps that cost a live session a whole batch of claims. Closes with a worked example, start to finish, on the question that produced all of it. Read hammer-coverage-navigate for routing and this for the route.\n\n skill: hammer-provider-quality\n uri: hammer://hammer-coverage-intelligence/skills/hammer-provider-quality\n Use when a provider-side quality or clinical director has a portfolio slot to fill and needs to know which improvement programme is worth it, whether a site really is worse than its peers, whether a before-and-after was the intervention or the mean coming back, or what a quality withhold and a savings model are actually worth to them. Carries the screen that runs before the pitch rather than after, as an ordered routing over estimators with nothing computed in this file; the margin rule that turns a boundary into a distance and says what the absence of a margin means; the redesign move each kind of death names, with the step it repairs, how many changes it takes, and the re-screen that has to clear before it is offered at all; a worked example quoted in bounds because bounds are what the receipts carry; and the refusal this seat has to publish rather than work around, because what an avoided admission is worth to you turns on a contract this world has not read. Read hammer-coverage-navigate for tool-level routing, hammer-building-the-evidence for the method underneath, and hammer-writing-the-deliverable when the answer has to become a document.\n\n skill: hammer-savings-classification\n uri: hammer://hammer-coverage-intelligence/skills/hammer-savings-classification\n Use when a saving has to be classified before it can be argued: whether a mechanism is hard or soft, which of the two savings methodologies a business case falls under, what evidence each is held to, and why a hard figure and a soft figure are never added together. Carries this world's definitions of hard and soft in full, the closed vocabulary of six mechanisms with each one's classification and the reason it follows, the shape of the ROI model each methodology implies, and the population chain that decides who books the saving. This is the source savings_methodology names in every payload it serves, so a reader holding a classification can check it here rather than taking it on trust. Read hammer-cost-of-care-case for the plan-side framing these classifications sit inside, and hammer-vendor-evidence if you are the one submitting the case.\n\n skill: hammer-vendor-evidence\n uri: hammer://hammer-coverage-intelligence/skills/hammer-vendor-evidence\n Use when a vendor, a digital health or device company, is building the case a health plan will read: what payers pay for the codes their product bills, which code a study should land in and what that code is worth, what their product would save a payer, what belongs in a business-case submission and what will get one declined. Opens with the translation problem, a vendor pitches clinical performance and a plan buys PMPM, and states the declined-submission failure mode plainly: a savings figure built from negotiated rates and presented as Cost of Care savings is the artefact the buyer's actuarial review exists to reject. Carries what this world can put into a submission with a receipt behind it, what it must refuse and why each refusal is worth more than the figure it withholds, and the pre-flight framing: what a draft is checked against and what gets named as a gap. Read hammer-coverage-navigate for tool-level routing; hammer-cost-of-care-case carries the buyer's own framing this file translates into.\n\n skill: hammer-writing-the-deliverable\n uri: hammer://hammer-coverage-intelligence/skills/hammer-writing-the-deliverable\n Use when the answer is not the deliverable and somebody needs the document: a commercial-terms report, a payer submission, a board memo, a benchmark pack. Carries how to declare the items the document needs before the first data call and read progress_check's verdict on the ledger against them, how to claim a computed figure, the section order and where each section's figures come from, the two appendices that make a document checkable, how a refusal becomes a section, how the claim report and deliver become the audit trail, and a worked example. Read after the persona skill that routes the question.\n\nWHEN YOU COME TO WRITE THE ANSWER UP, call say_world with the reader you are writing for. It changes how you say it and never what you may say.\n"
},
"kind": "served",
"licence": {
"coverage_database": "Contains information from the CMS Medicare Coverage Database.",
"cpt": "CPT is copyright the American Medical Association. Served under a distribution licence.",
"linkback": "https://hammer.ai/coverage-intelligence",
"terms": "https://hammer.ai/terms"
},
"payload": {
"ask_this_world_for": [
{
"answers": "What payers publish that they pay for a code, positioned against the public fee schedule that anchors it.",
"closed_vocabularies": [
{
"argument": "code",
"values_are_not_listed_here": "a vocabulary this long is read from this tool's own entry in `tools/list`, where every value is declared. It is summarised here because the whole of it would not fit in one tool result beside the rest of this orientation.",
"values_count": 213
},
{
"argument": "payer",
"values": [
"aetna",
"anthem",
"bcbs-mn",
"centene",
"cigna",
"hcsc",
"highmark",
"independence-blue-cross",
"unitedhealthcare",
"bcbs-north-carolina",
"uhc",
"ibx"
]
}
],
"declared_refusals": 8,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Rate position",
"tool": "rate_position"
},
{
"answers": "Two to four codes' payer rates set side by side under one receipt.",
"closed_vocabularies": [
{
"argument": "codes",
"values_are_not_listed_here": "a vocabulary this long is read from this tool's own entry in `tools/list`, where every value is declared. It is summarised here because the whole of it would not fit in one tool result beside the rest of this orientation.",
"values_count": 213
},
{
"argument": "payer",
"values": [
"aetna",
"anthem",
"bcbs-mn",
"centene",
"cigna",
"hcsc",
"highmark",
"independence-blue-cross",
"unitedhealthcare",
"bcbs-north-carolina",
"uhc",
"ibx"
]
}
],
"declared_refusals": 6,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Rate comparison",
"tool": "rate_comparison"
},
{
"answers": "Whether two or more codes may be billed together, per the NCCI PTP edit file.",
"closed_vocabularies": [
{
"argument": "codes",
"values_are_not_listed_here": "a vocabulary this long is read from this tool's own entry in `tools/list`, where every value is declared. It is summarised here because the whole of it would not fit in one tool result beside the rest of this orientation.",
"values_count": 213
}
],
"declared_refusals": 4,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Stack check",
"tool": "stack_check"
},
{
"answers": "How many people a screening programme would find positive and negative, true and false, and what its positive and negative predictive value would be, given a declared test sensitivity, specificity, prevalence and population size.",
"closed_vocabularies": [
{
"argument": "prevalence_basis",
"values": [
"administrative",
"measured",
"modelled",
"self_reported"
]
}
],
"declared_refusals": 5,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Screening yield",
"tool": "screening_yield"
},
{
"answers": "Whether a set of declared cost events are truly eliminated by an intervention or merely moved to a different time, reported as two separate totals that are never added together.",
"closed_vocabularies": [
{
"argument": "events.classification",
"values": [
"prevented",
"moved"
]
}
],
"declared_refusals": 2,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Deferred versus avoided",
"tool": "deferred_versus_avoided"
},
{
"answers": "Who a payer's own policy says may FURNISH and INTERPRET a code.",
"closed_vocabularies": [
{
"argument": "code",
"values_are_not_listed_here": "a vocabulary this long is read from this tool's own entry in `tools/list`, where every value is declared. It is summarised here because the whole of it would not fit in one tool result beside the rest of this orientation.",
"values_count": 213
}
],
"declared_refusals": 5,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Interpretation requirement",
"tool": "interpretation_requirement"
},
{
"answers": "How many people, and by what means somebody counted.",
"closed_vocabularies": [],
"declared_refusals": 14,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Cohort figures",
"tool": "cohort_figures"
},
{
"answers": "MOVED.",
"closed_vocabularies": [],
"declared_refusals": 1,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Comorbidity edges (moved)",
"tool": "comorbidity_edges"
},
{
"answers": "What the Medicare Physician Fee Schedule states for a code: whether the code splits into a professional component and a technical component, and the relative values it states for each.",
"closed_vocabularies": [
{
"argument": "code",
"values_are_not_listed_here": "a vocabulary this long is read from this tool's own entry in `tools/list`, where every value is declared. It is summarised here because the whole of it would not fit in one tool result beside the rest of this orientation.",
"values_count": 213
}
],
"declared_refusals": 5,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Fee components",
"tool": "fee_components"
},
{
"answers": "What a hospital filed as its own standard charge for an inpatient admission, under 45 CFR 180.",
"closed_vocabularies": [
{
"argument": "code_type",
"values": [
"MS-DRG",
"APR-DRG"
]
},
{
"argument": "filer",
"values": [
"cleveland-clinic",
"hca-medical-city",
"hss",
"mount-sinai",
"northwell",
"nyu-langone",
"vanderbilthealth-com"
]
},
{
"argument": "severity",
"values": [
"1",
"2",
"3",
"4"
]
}
],
"declared_refusals": 15,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Filed charge",
"tool": "filed_charge"
},
{
"answers": "MOVED, and it split in two, because it held two kinds of number.",
"closed_vocabularies": [],
"declared_refusals": 1,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Hypertension evidence (moved)",
"tool": "hypertension_evidence"
},
{
"answers": "One published observational association per answer: this exposure predicted that outcome, in people nobody randomised.",
"closed_vocabularies": [
{
"argument": "exposure",
"values": [
"masld_nafld",
"clinic_bp",
"nighttime_ambulatory_bp"
]
},
{
"argument": "outcome",
"values": [
"cardiovascular_event",
"incident_type_2_diabetes",
"liver_specific_mortality",
"all_cause_mortality",
"incident_heart_failure"
]
},
{
"argument": "stratum",
"values": [
"unstratified",
"higher_fibrosis"
]
}
],
"declared_refusals": 16,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Prognostic association",
"tool": "prognostic_association"
},
{
"answers": "One published RANDOMISED effect per answer: being assigned this arm changed that outcome's rate in a trial.",
"closed_vocabularies": [
{
"argument": "arm",
"values": [
"no_prior_cvd",
"prior_cvd",
"cardiac_rehabilitation"
]
},
{
"argument": "outcome",
"values": [
"incident_heart_failure",
"major_cardiovascular_event",
"all_cause_hospitalisation",
"all_cause_mortality"
]
}
],
"declared_refusals": 11,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Treatment effect",
"tool": "treatment_effect"
},
{
"answers": "What Galton's arithmetic predicts a cohort selected for being extreme will measure next period, before anybody intervened.",
"closed_vocabularies": [],
"declared_refusals": 4,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Regression to the mean expectation",
"tool": "regression_to_mean_expectation"
},
{
"answers": "How much of a readmission rate's movement between two periods is the survivor denominator moving underneath it.",
"closed_vocabularies": [],
"declared_refusals": 4,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Competing event coupling",
"tool": "competing_event_coupling"
},
{
"answers": "What a measure result is worth under a capitation withhold.",
"closed_vocabularies": [],
"declared_refusals": 4,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Withhold earn-back",
"tool": "withhold_earnback"
},
{
"answers": "Where a declared score sits against declared cut points, and how far the FIELD would have to move to cross each one with that score held still.",
"closed_vocabularies": [],
"declared_refusals": 5,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Tournament position",
"tool": "tournament_position"
},
{
"answers": "How much of a measured rate is the provider and how much is the sample it was measured on.",
"closed_vocabularies": [
{
"argument": "between_provider_variance.from",
"values": [
"hospital_variance"
]
}
],
"declared_refusals": 7,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Measure reliability",
"tool": "measure_reliability"
},
{
"answers": "Bring a customer's own aggregate figures into this world under a declared basis, and get back a handle the calculators read.",
"closed_vocabularies": [
{
"argument": "mapping.basis",
"values": [
"assumption",
"cited"
]
},
{
"argument": "mimetype",
"values": [
"application/json",
"text/csv"
]
}
],
"declared_refusals": 12,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Declared inputs",
"tool": "declared_inputs"
},
{
"answers": "Size an exposure over a declared population: what a condition accounts for across a book, over a declared period, on figures the caller declares.",
"closed_vocabularies": [],
"declared_refusals": 3,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Cost of care exposure",
"tool": "cost_of_care_exposure"
},
{
"answers": "Classify a declared savings mechanism as hard or soft against the buyer's own definitions, and say what evidence a case built on it is held to.",
"closed_vocabularies": [
{
"argument": "mechanism",
"values": [
"unit_cost_reduction",
"utilisation_reduction",
"demand_redirection",
"avoidance_of_future_cost",
"overpayment_recovery",
"adherence_improvement"
]
}
],
"declared_refusals": 2,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Savings methodology",
"tool": "savings_methodology"
},
{
"answers": "How far apart providers actually are on one published quality measure, read off a public measure corpus rather than declared by whoever needed the number.",
"closed_vocabularies": [
{
"argument": "measure_id",
"values": [
"COMP_HIP_KNEE",
"Hybrid_HWM",
"MORT_30_AMI",
"MORT_30_CABG",
"MORT_30_COPD",
"MORT_30_HF",
"MORT_30_PN",
"MORT_30_STK",
"PSI_03",
"PSI_04",
"PSI_06",
"PSI_08",
"PSI_09",
"PSI_10",
"PSI_11",
"PSI_12",
"PSI_13",
"PSI_14",
"PSI_15",
"PSI_90"
]
}
],
"declared_refusals": 8,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Hospital variance",
"tool": "hospital_variance"
},
{
"answers": "What one hospital filer carries, so a call that prices something is a call that answers.",
"closed_vocabularies": [
{
"argument": "filer",
"values": [
"cleveland-clinic",
"hca-medical-city",
"hss",
"mount-sinai",
"northwell",
"nyu-langone",
"vanderbilthealth-com"
]
}
],
"declared_refusals": 8,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Filer inventory",
"tool": "filer_inventory"
},
{
"answers": "The screen that runs before the pitch rather than after it: the caller's own declared candidate programmes, put through an ordered routing of steps, under one receipt.",
"closed_vocabularies": [
{
"argument": "candidates.mechanism",
"values": [
"unit_cost_reduction",
"utilisation_reduction",
"demand_redirection",
"avoidance_of_future_cost",
"overpayment_recovery",
"adherence_improvement"
]
},
{
"argument": "sweep.input",
"values": [
"between_provider_variance",
"numerator",
"denominator"
]
}
],
"declared_refusals": 6,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Candidate screen",
"tool": "candidate_screen"
},
{
"answers": "Where each part of a goal would be settled, before you spend a call on any of it.",
"closed_vocabularies": [
{
"argument": "rests_on.kind",
"values": [
"a_price_a_hospital_filed",
"a_rate_a_payer_published",
"what_one_filer_carries",
"a_published_effect_of_an_intervention",
"how_far_apart_providers_are_on_a_quality_measure",
"whether_a_difference_is_measurable_at_your_volume",
"how_many_patients_or_admissions",
"which_hospitals_are_one_system",
"what_your_own_contract_pays_you",
"what_something_costs_you_to_do",
"your_own_payer_mix_or_volume",
"whether_a_payer_paid_or_denied",
"whether_a_payer_covers_a_code",
"which_provider_is_cheaper_or_better",
"what_an_avoided_admission_is_worth_to_a_provider",
"whether_a_device_or_test_works"
]
}
],
"declared_refusals": 1,
"refusals_are_served_in": "tools/list, in this tool's own entry",
"title": "Where each part of a goal would be settled",
"tool": "frontier"
}
],
"cannot_answer": [
"a denial rate or a payment probability, from this corpus, by any method, including arithmetic on counts the world does serve; MRF files hold negotiated rates, not adjudications, so a rate's presence does not record a payment and its absence does not record a denial",
"whether a payer covers a code; a published negotiated rate says what a payer pays when it pays, not whether it will, so `payer X covers code Y` is not a sentence this corpus can settle",
"a cross-payer average without declared weights; the volume-weighted blend in the July rate report was licensed by a declared weighting, an unweighted national average is a number about this corpus's payer mix and not about any market, and the payer mix to weight by is the operator's to declare",
"a per-provider payment prediction or a ranking of providers; provider rows are observations of contracted rates, and 30 to 60 percent of small-cell variation is sampling error before anything else is said",
"a clinical claim about any device or test; this corpus prices procedures and holds no evidence about whether a product works, and a cited effect size for disease progression under an intervention class is never evidence about a named product's performance",
"what was in force as of a date beyond its own vintage; rates refresh monthly, LCDs change weekly, and upstream files are deleted at roughly 90 days, so the vintage a claim relied on has to be named explicitly, and a claim carried across vintages must be rechecked, never inherited",
"non-coverage from the absence of a rate row; payers omit, fragment and misfile disclosures market-wide, illustrated by one payer's own disclosure spanning roughly 11 million separate files and another payer's disclosure misusing the schema it is filed under, and the supportable sentence is that no rate was recorded in this month's disclosure, not that the code is not covered",
"a Medicare bundling verdict, and no bundling verdict at all for a pair with no edit row; the Medicare NCCI PTP baseline is gated behind an AMA CPT license acceptance form and is not held by this world, so the stacking edits served here are the Medicaid NCCI edition, which CMS publishes ungated, the two editions are not guaranteed to agree, and reporting a verdict from this corpus as `Medicare bundles these` states a rule this world has never read; separately, a pair with no row is UNKNOWN and never permission, since Category III codes routinely carry no row and 0689T against 91200 is exactly that case in the shipped file",
"a saving for a named payer's own book; the deferred-versus-avoided arithmetic runs over assumptions and cited transitions the caller declared, and a payer's realised saving depends on their population, their contracts and their utilisation, none of which this world holds, so quoting a modelled figure as that payer's saving states a fact about their business nobody here has measured",
"a total that sums a deferred cost with an avoided one; a cost that arrives later is not a cost that never arrives, and any figure adding a timing shift to a prevented event reports a saving that does not exist, which is why the two are returned as separate totals and no combined one is served at all",
"who may furnish or interpret a code, from the fee schedule; the PC/TC indicator and the status code say whether a code splits and whether Medicare prices it nationally, they do not say which specialty may furnish the professional component, and reading a restriction out of them would answer the most consequential half of that question with a field that does not contain it",
"how many people in a geography have MASLD, nor a prevalence converted between measurement classes; no source this world ships observes a liver below the national level, and administrative and epidemiological prevalence diverge in both directions by amounts that depend on the condition and the algorithm, so every cohort figure names the class it came from and this world holds no factor for converting between them",
"a causal effect from a comorbidity edge, a risk for the operator's own population, or a combined risk across more than one edge; every edge served here is observational and adjusted and adjustment is not randomisation, none of the cohorts behind them is a United States commercially insured screening population, and the severity-stratified estimates share their comparator with the unstratified ones so multiplying the two reports a hazard nobody measured",
"a causal effect from a hypertension prognostic record; nobody in those cohorts was randomised, adjustment is not randomisation, and a hazard ratio for a pressure predicting an outcome is not the effect of lowering that pressure",
"a combined figure across the hypertension corpus's two lists; a prognostic hazard ratio and a randomised risk ratio are different kinds of number measured against different comparators, so no arithmetic over one of each is a quantity anybody measured, and a caller who wants one of each asks twice and reads them side by side",
"a Hospital Readmissions Reduction Program penalty, an excess readmission ratio, or any 30-day readmission figure; no record this world holds counts 30-day unplanned readmissions for an index condition, and the nearest record it does hold, all-cause hospitalisation over 6 to 12 months in trial participants, has a different numerator, a different denominator and a different period",
"a HEDIS rate, a Star Rating, a quality bonus or a Medicaid capitation withhold; HEDIS measure specifications are NCQA's and licensed, the Star Ratings technical notes and the HRRP methodology are CMS's, none of the three is held here, so this world can report what the clinical literature published and never what any measure counts, which is the half a payment argument actually turns on",
"a cardiovascular risk or treatment effect for the operator's own population; every record names the people it was measured in, the cardiac rehabilitation trials recruited under 15 percent women, and the prognostic records are in people already being treated for hypertension",
"a hospital's charge compared with a payer's rate; a charge is what a provider filed and a rate is what a payer published, they are different kinds of money, and this world serves them through separate tools so that no single call can ask for one of each and no arithmetic can cross them",
"a hospital's filed payer name read as a payer; those names are free text each hospital wrote for itself, 240 distinct strings against the payer slugs the rate tools use, with one company appearing under several spellings, some entries naming a third-party administrator renting a network rather than the insurer, and some naming a government programme, so resolving one to a payer is an inference this world has not licensed",
"that a filer does not price an admission because its code is absent from their file; DRG groupers are revised and filers sit on different editions, so a missing code may be a code that edition does not define, and neither the file's schema version nor its vintage names the edition, which makes this the same refusal as declining to infer non-coverage from a missing rate row",
"a Cost of Care savings figure, from any money this world holds, by any method, including arithmetic done by hand; a plan's business-case standard scores savings on paid amounts, as opposed to allowed amounts or billed charges, and no compartment here produces a paid amount: payer MRFs hold negotiated rates, what a payer published it pays when it pays, and 45 CFR 180 filings hold filed charges, filed counterparty rates, filed brackets and cash prices, what a hospital published about its own prices, so a savings figure built from any of these and presented as Cost of Care savings is the artefact the buyer's actuarial review exists to reject, and what would settle it is adjudicated claims, which this world does not hold",
"that a modelled saving is incremental to a prior period's run rate, expressed on a plan-membership basis, or reported at gross value; the buyer's standard requires all three, and each is a fact about a plan's own book, its run rate, its membership and its accounting, which no corpus here observes, so all three enter a submission as the plan's own declared inputs and never as anything this world verified",
"a question from a branch of the medical-cost lever tree this world does not sit on; what payers publish that they pay is contracted-pricing evidence, so this world sits on the reduce-unit-costs branch and only its contracted-pricing leaf, and for every other branch the reason is one of three different facts rather than one word: a pharmacy question meets an absent corpus, since no pharmacy source is held here in any state and nobody has packed the data; an over-payment question meets an absent corpus and a licence wall together, since finding an overpayment needs the adjudications this world does not hold and the bundling-edit half of that work runs into the AMA CPT licence acceptance form the Medicare NCCI baseline sits behind; and a demand question, legitimate or not, meets a question this corpus cannot settle at all, since demand is utilisation, utilisation is adjudicated claims, and a price says nothing about how often anything happens; the branch is named so the question can be taken to a corpus that serves it rather than answered with an adjacent figure that happens to exist here",
"what an avoided admission or an avoided readmission is worth to the provider who avoided it; the same avoided admission is revenue that never arrives under fee for service and cost that is never incurred under a risk contract, so the sign of the figure is decided by the contract that provider is paid under, and this world has read no contract: it holds what payers published that they pay and what hospitals published that they charge, and neither records how any provider is paid for any population, so what would settle it is that provider's own contract terms and their own book, which are theirs to declare and never ours to infer from a published price"
],
"claim_discipline": "THE CLAIM DISCIPLINE. A world-backed answer ENDS IN A CLAIM REPORT: every sentence you wrote from this world, listed beside the verdict it got and the receipt it rests on. An answer without one is an answer whose claims were never put to anything.\nONLY `entailed` LICENSES WRITING A CLAIM DOWN. Every other verdict is a refusal with its own remedy, and neither checker falls back to the other: a checker that quietly fell back to a weaker one would be a fall-through approval with extra steps.\n`claim_check` DECIDES BY ARITHMETIC over the figures in the payload and never reads your sentence, so it is the same claim in any language. `judgement_check` READS THE ENGLISH and decides whether the sentence is one the answer supports. Choose by what you are actually checking, never by which one passed.\nQUOTING A FIGURE EXACTLY IS THE CASE THAT SEPARATES THEM. A decimal in a payload is read as rounded, so `== 1.43` against a payload carrying 1.43 is `not_entailed`: equality would assert a precision the answer never had. That verdict carries the two bounds that WOULD have passed, under `entailed_instead`. To put the published number into a SENTENCE, that is `judgement_check`'s question and it answers there.\nA FIGURE YOU COPY OUT OF ANOTHER ANSWER IS NOT CHECKED BY BEING TYPED IN. `claim_check` decides against one receipt and cannot see where a literal came from, so arithmetic mixing this answer's figure with a number carried over from another is your own arithmetic wearing a verdict. Where a world declares a figure uncombinable, that claim comes back `not_composable` and there is no better number to find: ask for each separately and read them side by side.\nAn answer taken and never checked is the one thing a check on submitted claims cannot see. `deliver` is what puts it in the audit log beside your report instead of nowhere.",
"how_to_read": {
"absences": [
{
"arose_as": "it is outside the vocabulary, so nobody ever looked and the absence is unreportable",
"refusal_kinds": [
"coverage/code-not-searched",
"coverage/outcome-not-in-corpus",
"coverage/stratum-not-in-corpus",
"coverage/measurement-not-in-corpus",
"coverage/arm-not-in-corpus",
"coverage/measure-not-in-corpus"
]
},
{
"arose_as": "this build never read it, whatever the source holds",
"refusal_kinds": [
"cohort/source-has-no-table",
"cohort/geography-not-built",
"cohort/period-not-built"
]
},
{
"arose_as": "the publisher held it back",
"refusal_kinds": [
"cohort/field-blocked",
"cohort/every-cell-withheld"
]
},
{
"arose_as": "somebody looked and the corpus holds nothing",
"refusal_kinds": [
"coverage/edge-not-measured",
"coverage/record-not-measured",
"coverage/cited-is-not-measured"
]
},
{
"arose_as": "the compartment was never mounted, so nothing could be read at all",
"refusal_kinds": [
"coverage/missing-data"
]
}
],
"basis_rule": "A figure computed from inputs of more than one basis carries the weakest basis among them, strongest to weakest being measured, then cited, then assumption, so a figure is never stronger than the weakest input that produced it.",
"dispersion": {
"note": "Two figures beside each cell say how thin that cell's own evidence is, and neither is a verdict. stddev_over_median is the cell's own stddev divided by its own median, both of them figures on this same cell: a value well above one says the spread dwarfs the middle, which is a property of the rows behind the cell rather than a finding about what the payer pays. largest_n_over_n is the largest row count any cell at or above the reporting floor carries for this code in this extract, divided by this cell's own row count, and largest_n_for_code names the payer and the count that division used so the arithmetic can be redone from this payload. No threshold is applied to either figure, no cell is withheld or refused on account of either, and this world states no opinion about which values make a cell worth standing behind: that reading is the caller's, on evidence this answer states rather than on prose it does not carry.",
"reported_by": [
"rate_position",
"rate_comparison"
]
},
"entailment": "`entailed` IS NECESSARY AND NOT SUFFICIENT. It is the only `claim_check` verdict that licenses writing a claim down, and it is not permission to conclude: it says the arithmetic over the figures in that one payload settles that one sentence. A claim resting on a figure whose value your own arguments carried in comes back refused rather than `entailed`, and every such figure is listed in the reply under `world.restates_your_input`. Every other verdict is a refusal with its own remedy, and `claim_check`'s entry in `tools/list` names each one, with what to do about it.",
"licensing_verdict": "entailed",
"refusal_is_a_value": "A REFUSAL IS A VALUE AND NOT AN ERROR. It comes back as an ordinary successful result carrying `kind: \"refusal\"`, a reason, and a `cannot_be_concluded` naming the sentence the evidence will not support. Retrying it returns the same refusal. Working around it by hand, having read it, is the one failure no check downstream can catch, because a claim you never submit is a claim nothing sees. The refusal is the finding: report it."
},
"kind": "orientation",
"navigation": {
"anchors": [
{
"anchor": "query",
"because": "a semantic index is present: text search over the indexed chunks answers",
"resolves": true
},
{
"anchor": "node",
"because": "a numbering is present: a canonical node id from an earlier answer resolves",
"resolves": true
},
{
"anchor": "select",
"because": "parquet tables attach: a table selector resolves and comes back with the matched rows' columns. `knowledge_navigate`'s own description lists the tables and their columns",
"resolves": true
}
],
"step_edges": {
"about_code": "claim → procedure or DRG",
"addresses": "article → procedure or DRG",
"asserts": "document → claim",
"bundled_with": "procedure (NCCI column 1) → procedure (column 2, the code denied)",
"carried_in": "register entry → data compartment",
"co_occurs_with": "condition → condition",
"co_priced": "procedure or DRG → procedure or DRG, stored both ways",
"computed_by": "tool → estimator",
"covers": "register entry → jurisdiction",
"declares": "register entry → column name",
"evidenced_by": "effect-size edge → citation",
"filed_charge": "hospital → procedure or DRG",
"files": "hospital → payer",
"gated_by": "register entry → access class",
"issued_by": "article → contractor",
"manual_asserts": "manual → claim",
"manual_issued_by": "manual → program or payer",
"manual_names": "claim → procedure or DRG",
"member_of": "procedure or DRG → code set",
"names": "refusal → data compartment",
"priced_by": "procedure → pricing class",
"raises_risk_of": "condition → outcome",
"rate_observed": "payer → procedure or DRG",
"read_by": "data compartment → tool",
"refuses": "tool → refusal",
"refuses_composition": "relation → corpus refusal",
"serves": "contractor → jurisdiction"
},
"steps": [
"carried_in",
"read_by",
"declares",
"refuses",
"names",
"gated_by",
"addresses",
"issued_by",
"bundled_with",
"priced_by",
"rate_observed",
"raises_risk_of",
"evidenced_by",
"co_occurs_with",
"covers",
"serves",
"computed_by",
"member_of",
"refuses_composition",
"asserts",
"about_code",
"co_priced",
"manual_names",
"manual_issued_by",
"manual_asserts",
"filed_charge",
"files"
]
},
"runtime_tools": [
{
"name": "judgement_check"
},
{
"name": "claim_check"
},
{
"name": "deliver"
},
{
"name": "session_trace"
},
{
"name": "progress_check"
},
{
"name": "model_advisory"
},
{
"name": "knowledge_navigate"
},
{
"name": "simulate"
},
{
"name": "sensitivity"
},
{
"name": "interval_check"
},
{
"name": "decision_navigate"
},
{
"name": "network_infer"
},
{
"name": "hello_world"
},
{
"name": "say_world"
}
],
"session": "49ec6090f32b112cd7fc2d39a72e0974",
"skills": [
{
"description": "Use when somebody has a commercial goal and needs evidence that survives the buyer's own scrutiny: pricing a service line, pitching a payer, sizing a market, arming a contract negotiation, showing an association's members what their data is worth. Carries the order of operations from a goal to a defensible claim, how to find what data exists before spending a call, the industry's own statistical toolkit mapped to the estimator that serves each, the confounds that have killed real findings in this corpus with what each one cost, how to read a refusal as the next step rather than a wall, and a worked example from a live commercial question to the evidence that answered it. This is the method; the persona skills route the question and hammer-writing-the-deliverable assembles the document.",
"mimeType": "text/markdown",
"name": "hammer-building-the-evidence",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-building-the-evidence"
},
{
"description": "Use when a plan-side Cost of Care analyst asks whether a use case is worth a slot, what a digital solution saves the plan, how to size an exposure over a declared population, whether a saving is hard or soft, what regression to the mean does to a selected cohort's next period, or anything phrased in PMPM, line of business, funding type or minimum impact threshold. Opens with the refusal that decides these sessions, the paid-amount wall: this world holds no paid amounts and cannot produce a Cost of Care savings figure that would pass the actuarial review a business case goes through. Carries the medical-cost lever tree with this world's position marked on it, the population chain from line of business down to funding type and why fully insured against administrative services only decides who carries the risk, which calculator serves which step of a business case, and the refusals to expect with what each one is telling you. Then the assembly order: name the branch, classify the mechanism before modelling it because hard and soft are different models rather than one model with a label, pin the population, size the savings side as a unit cost times a volume times a rate of change and say which of the three came from where, size the cost side that weak submissions omit (integration, stand-up, member outreach, recurring run cost), and put it on a clock, because a mid-year launch carries a full year of cost against a fraction of a year of savings. Ends on the rates that decide a soft case, which this world holds none of, and the rule that a readmission rate means nothing without its cause rule and time window. Read hammer-savings-classification for the hard and soft definitions in full, and hammer-coverage-navigate for tool-level routing; this is the plan-side framing over both.",
"mimeType": "text/markdown",
"name": "hammer-cost-of-care-case",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-cost-of-care-case"
},
{
"description": "Use when routing a coverage question to the tool that answers it, when you hold a GOAL rather than a question and do not yet know what to ask, or when deciding what this corpus does and does not license you to write. Opens with frontier, the tool that takes a goal in your own words plus the claims it rests on and routes each leaf to one of four places, settled here naming the tool, settled elsewhere naming a corpus in the register, yours to declare because your own book is a fact no corpus holds, or refused permanently together with the nearest question this world does answer. Then a routing table from what an operator is actually asking to the tool that serves it, covering payer rates (rate_position), the same rates for several codes under one receipt so a comparison can be checked (rate_comparison), code pairs against the Medicaid NCCI edit file (stack_check), who a payer's own policy says may furnish and interpret a code (interpretation_requirement), how the Medicare fee schedule splits a code into its professional and technical components (fee_components), what a hospital filed for an inpatient admission under 45 CFR 180, kept apart from payer rates because a charge is not a rate (filed_charge), headcounts and modelled prevalence with their measurement class (cohort_figures), published observational associations, MASLD or blood pressure, one record at a time (prognostic_association), randomised trial effects kept in a separate tool so no call can ask for one of each (treatment_effect), and the calculators that read no data at all (screening_yield, deferred_versus_avoided, regression_to_mean_expectation, competing_event_coupling). Explains the source register that knowledge_navigate reads, why an available source is different from an absent one, and why a register entry can never yield a figure. Names the refused questions with their cannot_answer entries. Covers the traps this corpus invites: percent-of-charge and phantom rows behind the quality filter, professional against technical against global, a rate is not coverage, cross-payer averaging without the operator's own payer mix, the two clocks, and the Medicaid-against-Medicare NCCI edition gap. Explains why positive predictive value collapses at low prevalence and why a deferred cost and an avoided cost are never summed. Every answer closes with the claim report defined in hammer-worlds.",
"mimeType": "text/markdown",
"name": "hammer-coverage-navigate",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-coverage-navigate"
},
{
"description": "Use when somebody asks what a named provider's contracts look like, whether an admission is priced as a DRG case rate or some other way, what a hospital filed for a procedure or an admission, or how a provider's filing compares with what payers publish. This is the working order for that question, one level above the tool surface: which question to settle first and why the obvious order is wrong, the arguments the tools require and the reason each exists, the refusals a real session hits and what each one is telling you, the crossing this world forbids together with the measured case where a careful analyst ran it and got a clean-looking answer that dissolved, how to read a row count in a corpus that replicates, and the checker traps that cost a live session a whole batch of claims. Closes with a worked example, start to finish, on the question that produced all of it. Read hammer-coverage-navigate for routing and this for the route.",
"mimeType": "text/markdown",
"name": "hammer-provider-contracting",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-provider-contracting"
},
{
"description": "Use when a provider-side quality or clinical director has a portfolio slot to fill and needs to know which improvement programme is worth it, whether a site really is worse than its peers, whether a before-and-after was the intervention or the mean coming back, or what a quality withhold and a savings model are actually worth to them. Carries the screen that runs before the pitch rather than after, as an ordered routing over estimators with nothing computed in this file; the margin rule that turns a boundary into a distance and says what the absence of a margin means; the redesign move each kind of death names, with the step it repairs, how many changes it takes, and the re-screen that has to clear before it is offered at all; a worked example quoted in bounds because bounds are what the receipts carry; and the refusal this seat has to publish rather than work around, because what an avoided admission is worth to you turns on a contract this world has not read. Read hammer-coverage-navigate for tool-level routing, hammer-building-the-evidence for the method underneath, and hammer-writing-the-deliverable when the answer has to become a document.",
"mimeType": "text/markdown",
"name": "hammer-provider-quality",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-provider-quality"
},
{
"description": "Use when a saving has to be classified before it can be argued: whether a mechanism is hard or soft, which of the two savings methodologies a business case falls under, what evidence each is held to, and why a hard figure and a soft figure are never added together. Carries this world's definitions of hard and soft in full, the closed vocabulary of six mechanisms with each one's classification and the reason it follows, the shape of the ROI model each methodology implies, and the population chain that decides who books the saving. This is the source savings_methodology names in every payload it serves, so a reader holding a classification can check it here rather than taking it on trust. Read hammer-cost-of-care-case for the plan-side framing these classifications sit inside, and hammer-vendor-evidence if you are the one submitting the case.",
"mimeType": "text/markdown",
"name": "hammer-savings-classification",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-savings-classification"
},
{
"description": "Use when a vendor, a digital health or device company, is building the case a health plan will read: what payers pay for the codes their product bills, which code a study should land in and what that code is worth, what their product would save a payer, what belongs in a business-case submission and what will get one declined. Opens with the translation problem, a vendor pitches clinical performance and a plan buys PMPM, and states the declined-submission failure mode plainly: a savings figure built from negotiated rates and presented as Cost of Care savings is the artefact the buyer's actuarial review exists to reject. Carries what this world can put into a submission with a receipt behind it, what it must refuse and why each refusal is worth more than the figure it withholds, and the pre-flight framing: what a draft is checked against and what gets named as a gap. Read hammer-coverage-navigate for tool-level routing; hammer-cost-of-care-case carries the buyer's own framing this file translates into.",
"mimeType": "text/markdown",
"name": "hammer-vendor-evidence",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-vendor-evidence"
},
{
"description": "Use when the answer is not the deliverable and somebody needs the document: a commercial-terms report, a payer submission, a board memo, a benchmark pack. Carries how to declare the items the document needs before the first data call and read progress_check's verdict on the ledger against them, how to claim a computed figure, the section order and where each section's figures come from, the two appendices that make a document checkable, how a refusal becomes a section, how the claim report and deliver become the audit trail, and a worked example. Read after the persona skill that routes the question.",
"mimeType": "text/markdown",
"name": "hammer-writing-the-deliverable",
"uri": "hammer://hammer-coverage-intelligence/skills/hammer-writing-the-deliverable"
}
],
"world": {
"build": 2,
"description": "What payers publish that they pay, code by code, and what may be said with it.",
"name": "hammer-coverage-intelligence",
"receipt_is_redeemed_against": "this world's name, vintage and build together; a receipt naming a different vintage is a different world's",
"vintage": "2026-10-01"
},
"worlds_do_not_compose": "TWO WORLDS' ANSWERS DO NOT COMPOSE INTO A THIRD CLAIM. Every answer is redeemed against one world's receipt and THERE IS NO RECEIPT SPANNING TWO, so a sentence built from this world and another is a claim neither world issued and neither checker can check. Report the two findings separately, each against its own receipt, and leave the join to the reader as the reader's own inference."
},
"receipt": null,
"session": "49ec6090f32b112cd7fc2d39a72e0974",
"tool": "hello_world",
"world": {
"build": 2,
"description": "What payers publish that they pay, code by code, and what may be said with it.",
"name": "hammer-coverage-intelligence",
"vintage": "2026-10-01"
}
}Permanent saved result
claim_check
judgement_check