Verdict Engine - Live Coverage

148 checks. 15 categories. All running.

This page reads directly from the running Verdict Engine (grants-verdict-engine:8981/v1/catalog). The numbers below are not marketing - they are a live read.

148
Total checks defined
120
Active at pre-clearance (application stage)
28
Active during grant lifecycle (post-approval)

Category map

Each category contributes one mean score to the aggregate. Weights below sum to 1.0. Lifecycle-stage-gated categories run their checks at the moment they apply; at pre-clearance they emit a documented skipped_reason: stage_gatedrather than a fake pass.

CodeCategoryChecksWeightStageScope
GOVGovernance1410%Pre-clearanceBoard structure, conflicts, founding instruments, charter scope, governance rules.
FINFinancial1410%Pre-clearanceAudited financials, reserve adequacy, restricted-fund segregation, related-party transactions. PDF extraction via pdf-parse on Form 990 and audited financials.
SANSanctions1215%Pre-clearanceOFAC SDN, EU consolidated, UK HMT, UN Security Council, MAS targeted, AUSTRAC, secondary-sanctions exposure. Backed by sanctions-screening-cache.
BENBeneficial Ownership88%Pre-clearanceNatural-person identification, control chains, PEP screening, address and ID verification. Backed by kyc-service.
PRGProgrammatic116%Pre-clearanceTheory of change coherence, prior-outcome verification, beneficiary methodology, equivalency-determination evidence. Optional Bedrock LLM 60/40 blend with heuristic NLP scorers.
COMCompliance106%Pre-clearance18 attestation types - charitable status, sanctions self-attest, AML, FATCA/CRS, anti-bribery, conflicts policy.
OPSOperational95%Pre-clearanceStaff capacity, prior grant lifecycle, complaint history, organizational age, audit infrastructure.
MILMilestone Evidence127%LifecycleMilestone definition quality, evidence pre-positioning, on-time submission rate, milestone-disbursement linkage. Active during grant lifecycle (post-approval).
DISDisbursement95%LifecycleBank-account verification, disbursement-trace auditability, no-co-signer divergence, multi-signature controls. Active at disbursement events.
PERPersonnel84%Pre-clearanceKey-person continuity, criminal-record disclosure, regulatory bar history, board-of-directors current standing.
RPTReporting74%LifecycleOutcome-report quality, tax-filing completeness, regulator-required submission cadence, narrative-financial reconciliation. Active during reporting cadence.
EQUEquivalency75%Pre-clearanceIRC 4945 ED evidence, ER fallback adequacy, currency-conversion documentation, tax-treaty applicability.
JURJurisdictional95%Pre-clearanceCharity Commission match, IRS Tax-Exempt status, ACRA, DIFC, RFB, ESA Stiftungsverzeichnis.
NETNetwork Effects94%Pre-clearanceInter-foundation grant-history overlap, beneficial-owner overlap, prior-incident registry match, related-entity flags.
ANOAnomaly Detection96%Pre-clearanceSubmission-pattern anomalies, rate-of-applications, document-metadata anomalies, suspected templating.
Total15 categories148100%Aggregate score = weighted mean of category means

What "running" means

Every check is implemented. None are stubs returning a constant. Each check returns one of:

  • PASS / FAILwith score and detail object containing the evidence
  • CONDITIONALwith reviewable caveats
  • SKIPPEDwith an explicitskipped_reasonwhen context is missing or the check is stage-gated. Skips are documented, not silent passes.

If a category looks low-coverage on a particular verdict, the per-check skipped_reasontells the foundation exactly what evidence to add. The platform does not soft-pass an absent input.

Catalog version

Current catalog version:2026.05.04-148. Version changes are CourtChain-anchored along with a diff manifest so foundations can audit what changed between runs.

Engagement materialsinclude the full check-by-check reference - what each ID evaluates, the primary signal source, the scoring band, and the evidence the engine consumes.Request the engagement pack →