{
  "$schema": "https://ainumbers.co/schema/ap2-templates-v1.1.json",
  "registry_version": "1.1.0",
  "updated": "2026-06-12",
  "maintainer": "Post Oak Labs",
  "license": "CC BY 4.0",
  "description": "AINumbers ChainGraph AP2 template registry. One entry per mandate_type. Describes regulatory citations, primary tasks, target audiences, stochastic framing rules, escalation conditions, and the prompt template_body used to translate a ChainGraph execution artifact into a regulator-grade LLM prompt. Internal mandate_type taxonomy — not AP2 v0.2 FIDO Alliance spec vocabulary.",
  "separately_licensable": true,
  "licensing_contact": "tim@postoaklabs.com",
  "llmeo_note": "This registry is the data layer for PTG-01 (compose_ap2_prompt) and is the primary LLMEO surface for compliance orchestrators. Each template_body is designed to be injected into an LLM system prompt alongside a ChainGraph artifact JSON, producing a regulator-grade narrative anchored to the execution_hash.",
  "template_body_usage": "In template_body, double-curly-brace tokens (e.g. {{execution_hash}}) are populated by PTG-01 from the source artifact. The execution_hash anchor footer is mandatory in every narrative — it is the audit traceability claim that makes ChainGraph outputs verifiable.",
  "templates": {
    "liquidity_mandate": {
      "display_name": "Liquidity Mandate (LCR/NSFR)",
      "tool_ids": ["sim-01-lcr-nsfr-liquidity-stress-test"],
      "regulatory_citations": [
        "Basel III LCR — CRR Art. 412 (minimum 100%)",
        "Basel III NSFR — CRR Art. 428 (minimum 100%)",
        "EBA GL/2017/01 (LCR supervisory reporting)",
        "ECB/2016/17 (SSM LCR supervisory expectations)",
        "BCBS 238 (LCR final rule)"
      ],
      "primary_tasks": ["board_memo", "regulatory_submission", "risk_memo", "plain_english_summary"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "Report percentiles as 'the P5 breach probability is X%', not 'the probability of breach is X%'",
        "Never present a single simulation path as 'the' outcome — always cite the percentile",
        "When citing time-to-breach, state the percentile: 'P50 time-to-breach is N days'",
        "State the simulation path count: 'across {{policy_parameters.num_paths}} Monte Carlo paths'"
      ],
      "escalation_conditions": [
        "LCR P5 < 100% — regulatory minimum breach risk",
        "NSFR P5 < 100%",
        "P50 time-to-breach < 30 days",
        "Stress scenario delta > 15 percentage points vs. base"
      ],
      "audience_formats": {
        "board": "Executive summary (3 bullets max), traffic-light verdict (Pass/Warn/Breach), recommended board action",
        "risk_committee": "Full percentile table (P5/P25/P50/P75/P95), sensitivity tornado, breach-trigger analysis, escalation recommendation",
        "regulator": "Structured regulatory submission format, cite CRR articles, quantified shortfall if any, remediation timeline",
        "quant": "Full statistical output with methodology footnotes, model assumptions, Monte Carlo parameters, convergence diagnostics",
        "ops": "Operational action items, buffer top-up instructions, reporting calendar, responsible owner per action"
      },
      "tone_modifiers": {
        "formal": "Use passive voice where appropriate. Avoid contractions. Lead with the regulatory determination.",
        "technical": "Include model parameters and statistical methodology. Reference specific article numbers.",
        "plain": "Explain what the numbers mean in practical terms. Avoid regulatory jargon unless unavoidable."
      },
      "template_body": "You are a treasury risk officer reviewing a Basel III liquidity stress simulation (execution_hash: {{execution_hash}}, tool: {{tool_id}}, N={{policy_parameters.num_paths}} Monte Carlo paths × {{policy_parameters.num_steps}} time steps).\n\nKey results:\n- LCR at P50: {{output_payload.lcr_p50}}% | P5: {{output_payload.lcr_p5}}% | P95: {{output_payload.lcr_p95}}%\n- NSFR at P50: {{output_payload.nsfr_p50}}% | P5: {{output_payload.nsfr_p5}}% | P95: {{output_payload.nsfr_p95}}%\n- LCR breach probability: {{output_payload.lcr_breach_prob}}%\n- Median time-to-breach: {{output_payload.median_ttb}} days\n- Top sensitivity driver: {{output_payload.top_sensitivity_driver}}\n\nYour task: write an ALCO-ready liquidity risk memo (max 400 words). Structure: (1) Regulatory compliance status (cite CRR Art. 412/428), (2) Breach probability interpretation, (3) Top sensitivity findings, (4) Recommended management actions with urgency flags.\n\nNever present a single simulation path as 'the' outcome — always cite the percentile band.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "payment_mandate": {
      "display_name": "Payment Mandate (AP2 / ACP / Agentic Commerce)",
      "tool_ids": ["art-01-ap2-mandate-chain-validator", "art-12-acp-checkout-conformance-validator"],
      "regulatory_citations": [
        "AP2 v0.2 — FIDO Alliance (April 2026) — Intent/Cart/Payment mandate schema",
        "OpenAI/Stripe ACP (Agentic Commerce Protocol, 2026) — CheckoutRequest/Response schema",
        "PSD2 Art. 4(5) (payment mandate definitions)",
        "PSD3 (proposed) — AI-initiated payment agent obligations",
        "EU Instant Payments Regulation (IPR) — in force October 2025"
      ],
      "primary_tasks": ["risk_memo", "regulatory_submission", "plain_english_summary", "remediation_plan"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "Payment mandate verdicts are DETERMINISTIC — never frame as probabilistic",
        "Use CONFORMANT / NON-CONFORMANT / PARTIAL verdicts only — no hedged language",
        "If over-spend is detected, state the exact delta in the mandate's declared currency",
        "Cite the specific AP2 or ACP clause violated, not a general statement"
      ],
      "escalation_conditions": [
        "Signature-chain validation failure — suspend mandate immediately",
        "Scope/limit mismatch between Intent and Cart mandates",
        "TTL/expiry violation — mandate must be re-issued",
        "Over-spend detection trigger — block settlement pending review",
        "Human-Not-Present flow on mandate > €1,000 — require explicit human authorisation",
        "ACP token signature prefix invalid — reject checkout, log for security review"
      ],
      "audience_formats": {
        "board": "Strategic summary of agentic payment risk posture, key failure modes, recommended governance controls",
        "risk_committee": "Technical failure analysis, failing predicates enumerated, risk scoring per category, remediation priorities",
        "regulator": "AP2 v0.2 conformance status per FIDO specification section, failing clauses cited, remediation plan",
        "quant": "Detailed validation output, predicate-by-predicate breakdown, test vector analysis",
        "ops": "Integration checklist, failing fields with fix instructions, testing protocol, go-live readiness score"
      },
      "tone_modifiers": {
        "formal": "Lead with the conformance verdict. Enumerate failures in structured list format.",
        "technical": "Reference AP2 v0.2 specification sections and ACP schema definitions. Include field-level analysis.",
        "plain": "Describe what each failure means for a user trying to make a payment."
      },
      "template_body": "You are a payment compliance officer reviewing a ChainGraph payment mandate audit artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nVerdict summary:\n- Overall conformance: {{output_payload.overall_verdict}}\n- AP2 mandate chain: {{output_payload.mandate_chain_status}}\n- Violations: {{output_payload.violations}}\n- ACP conformance (if present): {{output_payload.acp_verdict}}\n- compliance_flags: {{compliance_flags}}\n\nYour task: draft a one-page payment mandate audit memo. Structure: (1) Executive verdict, (2) Clause-level findings (cite AP2/ACP clause numbers — AP2 v0.2 §section or ACP field name), (3) Required remediation steps, (4) Audit anchor.\n\nDo NOT frame deterministic verdicts as probabilistic. Use CONFORMANT / NON-CONFORMANT / PARTIAL only.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "payment_policy": {
      "display_name": "Payment Policy (Agent Spend)",
      "tool_ids": ["art-02-agent-spend-policy-simulator"],
      "regulatory_citations": [
        "AP2 v0.2 Human-Not-Present flows — FIDO Alliance",
        "PSD2 Article 75 (Unauthorized Payment Transactions)",
        "PSD2 Article 10 (spending limit controls for payment initiation)",
        "FCA Consumer Duty — PRIN 2A (July 2023)",
        "CFPB Open Banking Rule §1033 (agent-initiated payment data access)"
      ],
      "primary_tasks": ["risk_memo", "remediation_plan", "top_3_risks", "board_memo"],
      "audiences": ["board", "risk_committee", "quant", "ops"],
      "stochastic_framing_rules": [
        "Simulation outputs are synthetic — label transaction paths as 'simulated across {{policy_parameters.num_transactions}} synthetic transactions'",
        "Cite the simulation seed and transaction count so results are reproducible",
        "P5/P95 intervals must be reported alongside median — never report a single point estimate for breach rates",
        "Policy-bypass paths are illustrative, not exhaustive — state this clearly"
      ],
      "escalation_conditions": [
        "P95 breach rate > 5% — escalate to risk committee before agent deployment",
        "Human-Not-Present bypass path detected — mandatory human-in-the-loop gate required",
        "Cumulative ceiling breach detected in > 1% of simulated trajectories — tighten velocity controls",
        "Autonomous agent scope creep beyond approved categories confirmed"
      ],
      "audience_formats": {
        "board": "Policy strength verdict, top 3 bypass risk categories, recommended policy tightening actions",
        "risk_committee": "Full bypass-path enumeration, risk-scored by likelihood and impact, policy gap matrix",
        "quant": "Transaction-path statistics, policy coverage rate, false-negative analysis methodology",
        "ops": "Policy rule corrections, integration test cases for each bypass path, deployment checklist"
      },
      "tone_modifiers": {
        "formal": "Frame as a policy control assessment. Use risk management vocabulary.",
        "technical": "Detail the transaction parameters that trigger each bypass. Include rule logic.",
        "plain": "Explain what a bad actor or misconfigured agent could do under the current policy."
      },
      "template_body": "You are a treasury risk analyst reviewing an agent spend-policy simulation (execution_hash: {{execution_hash}}, tool: {{tool_id}}, N={{policy_parameters.num_transactions}} synthetic transactions, seed={{policy_parameters.seed}}).\n\nKey simulation results:\n- Median breach rate: {{output_payload.median_breach_rate}}%\n- P95 breach rate: {{output_payload.p95_breach_rate}}%\n- Human-Not-Present bypass paths: {{output_payload.hnp_bypass_count}}\n- Top violation category: {{output_payload.top_violation_category}}\n- compliance_flags: {{compliance_flags}}\n\nYour task: write an executive risk memo (max 300 words). Structure: (1) Policy health verdict, (2) Top three breach scenarios with AP2 clause citations, (3) Recommended policy tightening steps, (4) Board escalation recommendation if P95 > 5%.\n\nAlways prefix simulation results with 'Simulated estimate:' — never 'The system concludes'. Report P5/P95 intervals, not single point estimates.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "compliance_mandate": {
      "display_name": "Compliance Mandate",
      "tool_ids": [
        "art-04-agent-identity-attestation-checker",
        "art-08-en16931-einvoice-batch-validator",
        "art-11-vop-batch-match-rate-analyser",
        "art-12-acp-checkout-conformance-validator",
        "art-13-eudi-wallet-credential-readiness-checker",
        "cry-01-zk-compliance-proof-generator",
        "rca-03-iso20022-address-migration-verifier"
      ],
      "regulatory_citations": [
        "FATF Travel Rule (Recommendation 16)",
        "EU AMLR 2024/1624 — full application July 2027 / AMLA operations 2028",
        "EU Instant Payments Regulation VoP — in force October 2025",
        "SWIFT CBPR+ structured-address mandate — November 2026",
        "EN 16931-1:2017 / EU ViDA — structured e-invoicing mandate (France large/medium Sep 2026)",
        "GDPR Article 25 (Privacy by Design and by Default)",
        "eIDAS 2.0 / EUDI Wallet ARF v1.4 — member-state rollout November 2026",
        "EU AI Act (2024/1689) Annex III — high-risk AI system obligations"
      ],
      "primary_tasks": ["regulatory_submission", "risk_memo", "remediation_plan", "board_memo", "top_3_risks"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "Field-level validation verdicts are DETERMINISTIC — use PASS/FAIL/PARTIAL only",
        "Match-rate outputs (VoP, address validation) are statistical — cite sample size and confidence interval",
        "ZK proof tokens (CRY-01) are synthetic and explicitly educational — always include the disclaimer in any narrative",
        "Countdown to hard regulatory deadlines must be derived from today's date, not hardcoded"
      ],
      "escalation_conditions": [
        "Compliance pass rate < 95% — escalate to operations before reprocessing",
        "Deadline is within 90 days of assessment date",
        "Credential chain depth > 4 hops (ART-04) — security architecture review required",
        "EUDI readiness score < 60% with < 90 days to deadline — immediate compliance programme escalation",
        "ISO 20022 batch readiness < 70% with < 60 days to SWIFT CBPR+ mandate — emergency migration track",
        "ZK proof token used outside educational context — legal review required"
      ],
      "audience_formats": {
        "board": "Compliance readiness verdict (Ready / Partially Ready / Not Ready), deadline exposure summary, board resolution template",
        "risk_committee": "Field-level compliance breakdown, deadline risk matrix, remediation timeline with owners",
        "regulator": "Structured compliance statement per applicable regulation and article, quantified non-conformance, remediation commitment",
        "quant": "Statistical batch analysis, field-error distribution, country-specific breakdown, false-positive/negative rates",
        "ops": "Field-by-field fix list, tooling recommendations, testing protocol, cutover checklist"
      },
      "tone_modifiers": {
        "formal": "Lead with the compliance determination. Enumerate non-conformances in structured list.",
        "technical": "Cite specific regulatory article numbers. Include field-name references.",
        "plain": "Explain what non-compliance means in practice and what needs to change."
      },
      "template_body": "You are a compliance analyst reviewing a ChainGraph compliance mandate artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nCompliance summary:\n- Mandate: {{mandate_type}} / {{tool_id}}\n- Overall verdict: {{output_payload.verdict}}\n- Failure count: {{output_payload.failure_count}}\n- Key compliance flags: {{compliance_flags}}\n- Days to nearest hard deadline: {{output_payload.days_to_deadline}}\n\nYour task: produce a gap analysis memo (max 400 words). Structure: (1) Regulatory context — cite the specific article/directive and its deadline, (2) Verdict with failure breakdown, (3) Prioritised remediation roadmap (deadline-weighted), (4) Escalation recommendation.\n\nField-level verdicts are DETERMINISTIC — use PASS/FAIL/PARTIAL only. Do not hedge deterministic findings. Countdown to deadline must use actual days remaining.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "capital_assessment": {
      "display_name": "Capital Assessment (Basel / RWA)",
      "tool_ids": ["art-07-basel31-reporting-delta-calculator", "sim-03-basel-rwa-scenario-modeler"],
      "regulatory_citations": [
        "Basel 3.1 / BCBS CRE20–CRE99 (standardised approach credit risk)",
        "BCBS CAP30 — output floor: 72.5% of SA RWA",
        "UK PRA PS1/26 — Basel 3.1 effective January 1, 2027",
        "EU CRR3 (Capital Requirements Regulation 3, 2025)",
        "BCBS d424 (IRB capital formula — Φ⁻¹ rational approximation)",
        "FRTB MAR33 (Market Risk — Internal Models Approach)"
      ],
      "primary_tasks": ["board_memo", "regulatory_submission", "risk_memo", "period_comparison"],
      "audiences": ["board", "risk_committee", "regulator", "quant"],
      "stochastic_framing_rules": [
        "Deterministic delta calculations (ART-07) — use exact basis-point values, never ranges",
        "Monte Carlo scenario outputs (SIM-03) — always report P5/P50/P95 with simulation parameters",
        "Output floor binding is deterministic once SA RWA is computed — frame as a compliance test, not a probability",
        "Capital shortfall figures are point-in-time estimates — note the reference date from policy_parameters",
        "Label all RWA computations as 'model estimate, not regulatory determination'"
      ],
      "escalation_conditions": [
        "CET1 drops below 12% in any scenario — immediate capital planning escalation",
        "Output floor becomes binding (SA RWA > IRB RWA) — A-IRB model utility review required",
        "Capital requirement increase > 20% vs. current approach — material impact ICAAP review",
        "P5 RWA scenario exceeds current capital buffer — stress buffer adequacy review"
      ],
      "audience_formats": {
        "board": "Capital impact summary (£/$ amount + %, traffic-light vs. targets), output-floor status, strategic options",
        "risk_committee": "RWA delta table by exposure class, floor binding analysis, sensitivity to key parameters, remediation levers",
        "regulator": "Basel 3.1 compliance statement per PS1/26 requirements, quantified impact by approach, implementation timeline",
        "quant": "Full scenario model output, approach comparisons, Cholesky/correlation assumptions, convergence notes"
      },
      "tone_modifiers": {
        "formal": "Lead with the capital impact determination. Reference specific Basel articles.",
        "technical": "Include model assumptions, SA-CR / IRB approach parameters, output-floor calculation.",
        "plain": "Explain how much more capital is required and why."
      },
      "template_body": "You are a capital adequacy analyst reviewing a ChainGraph Basel 3.1 capital assessment (execution_hash: {{execution_hash}}, tools: art-07 + sim-03).\n\nCapital assessment summary:\n- Current CET1: {{output_payload.current_cet1}}% | Post-Basel 3.1: {{output_payload.post_b31_cet1}}%\n- CET1 delta: {{output_payload.cet1_delta}} bps\n- Output floor binding: {{output_payload.floor_binds}} (72.5% of SA RWA = {{output_payload.floor_rwa}})\n- Capital shortfall vs 12.5% total requirement: {{output_payload.capital_shortfall}}\n- Highest RWA delta asset class: {{output_payload.top_delta_asset_class}} (+{{output_payload.top_delta_bps}} bps)\n\nYour task: write an ICAAP-ready capital adequacy memo (max 500 words). Structure: (1) Regulatory context (cite BCBS CAP30 and UK PRA PS1/26, deadline 2027-01-01), (2) RWA delta analysis by asset class, (3) Output floor impact, (4) CET1 trajectory and remediation options, (5) Board escalation flag if shortfall detected.\n\nDeterministic delta results use exact values. Monte Carlo scenario results cite P5/P50/P95. Never present model output as a regulatory determination.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "risk_control": {
      "display_name": "Risk Control (AML / Market / Consumer)",
      "tool_ids": ["art-10-amla-transaction-typology-risk-scorer", "qfa-02-portfolio-var-engine"],
      "regulatory_citations": [
        "FATF AML Typologies (2024)",
        "FATF Recommendation 15 (new technologies — AI-assisted risk assessment)",
        "FATF Travel Rule (Recommendation 16)",
        "EU AMLR 2024/1624 Art. 19 (transaction monitoring obligations) — July 2027",
        "Basel III FRTB (BCBS d352/d457) — IMA VaR/ES requirements",
        "ESMA Guidelines on performance scenarios (PRIIPs KID)",
        "FCA Consumer Duty — PRIN 2A (July 2023)",
        "UK PSR — APP Reimbursement (October 2024)"
      ],
      "primary_tasks": ["risk_memo", "top_3_risks", "board_memo", "remediation_plan"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "AML typology scores are synthetic simulation outputs — label as 'indicative typology risk', never 'confirmed AML'",
        "Never identify a specific transaction or entity as fraudulent — simulation is educational only",
        "VaR and ES are model-dependent estimates — always cite the method (Historical/Parametric/Monte Carlo) and confidence level",
        "AML risk scores feed SAR triage pre-screening only — human investigator review is mandatory before filing"
      ],
      "escalation_conditions": [
        "AML risk score ≥ HIGH on any account cluster — mandatory human investigator review",
        "FATF Travel Rule predicate triggered (amount ≥ $1,000 threshold)",
        "VaR breach in back-test > 5 exceedances in 250 days — FRTB capital model review required",
        "Expected Shortfall at 97.5% exceeds stressed VaR — risk model audit",
        "APP fraud graph cluster score above threshold",
        "Consumer Duty outcome metric below benchmark"
      ],
      "audience_formats": {
        "board": "Risk posture summary (High/Medium/Low), key typology exposures, governance recommendations",
        "risk_committee": "Typology-by-typology scoring, cluster analysis summary, VaR methodology and results, threshold calibration",
        "regulator": "AML control framework statement, typology coverage, Travel Rule implementation status",
        "quant": "Full graph analysis output, scoring methodology, VaR parameters, false-positive/negative calibration",
        "ops": "Account/cluster action list, investigation workflow, escalation triggers, reporting obligations"
      },
      "tone_modifiers": {
        "formal": "Use risk management vocabulary. Lead with the overall risk determination.",
        "technical": "Include typology codes, graph metrics, VaR parameters, threshold definitions.",
        "plain": "Explain what each risk indicator means and what action it requires."
      },
      "template_body": "You are a risk officer reviewing a ChainGraph risk control artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nRisk summary:\n- Risk determination: {{output_payload.risk_determination}}\n- Key findings: {{output_payload.key_findings}}\n- Compliance flags: {{compliance_flags}}\n\nFor AML tools (ART-10): Draft a SAR triage pre-screening memo citing FATF Recommendation 15 and EU AMLR Art. 19. CRITICAL: label all outputs as 'indicative typology risk' — this is a model output; human investigator review is mandatory before any SAR filing.\n\nFor VaR tools (QFA-02): Draft a risk management report noting method ({{policy_parameters.var_method}}), confidence level ({{policy_parameters.confidence_level}}%), and any back-test exceedances. Report P5/P50/P95 ES alongside VaR — never a single point estimate.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "model_governance": {
      "display_name": "Model Governance (EU AI Act)",
      "tool_ids": ["art-05-eu-ai-act-credit-scoring-conformity"],
      "regulatory_citations": [
        "EU AI Act (2024/1689) Annex III Part 5(b) — credit-scoring high-risk (fully applicable August 2, 2026)",
        "EU AI Act Art. 9 (risk management system requirements)",
        "EU AI Act Art. 10 (data governance requirements)",
        "EU AI Act Art. 11 (technical documentation — Annex IV skeleton)",
        "EU AI Act Art. 13 (transparency and provision of information to users)",
        "EU AI Act Art. 14 (human oversight requirements)",
        "EU AI Act Art. 15 (accuracy, robustness and cybersecurity)",
        "EBA GL/2020/06 (model risk management — supervisory expectations)"
      ],
      "primary_tasks": ["regulatory_submission", "risk_memo", "remediation_plan", "board_memo"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "Disparate impact ratios and equalized odds gaps are deterministic calculations given the input dataset statistics",
        "Bias metrics are computed on synthetic data — label all test results as 'indicative, not production'",
        "Conformity self-assessment is a structured determination, not a legal opinion — flag for qualified counsel review",
        "Hard deadline August 2, 2026 is absolute — state countdown in days, not quarters"
      ],
      "escalation_conditions": [
        "Disparate impact ratio < 0.80 (4/5ths rule) on any protected characteristic — mandatory bias remediation before deployment",
        "Equalized odds gap > 5 percentage points — model risk committee review, Article 14 human oversight trigger",
        "Data quality attestation fails completeness check — training data remediation before conformity assessment",
        "< 30 days to August 2, 2026 with non-conformant status — emergency compliance programme activation",
        "Technical documentation skeleton incomplete — regulatory filing blocked"
      ],
      "audience_formats": {
        "board": "AI Act compliance status (August 2026 deadline), failing dimensions, legal exposure summary, recommended board action",
        "risk_committee": "Full bias test results by characteristic, documentation gap analysis, remediation roadmap",
        "regulator": "Article 11 technical documentation in prescribed format, conformity self-assessment, bias testing methodology and results",
        "quant": "Statistical bias test output, disparate impact methodology, equalized odds computation, data quality attestation detail",
        "ops": "Documentation checklist, data pipeline requirements, monitoring obligations, post-deployment reporting cadence"
      },
      "tone_modifiers": {
        "formal": "Lead with the conformity determination. Reference EU AI Act articles directly.",
        "technical": "Include statistical test parameters, threshold definitions, methodology citations.",
        "plain": "Explain what the AI Act requires and where the model falls short."
      },
      "template_body": "You are an AI governance officer reviewing a ChainGraph EU AI Act conformity assessment (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nConformity summary:\n- Overall verdict: {{output_payload.conformity_verdict}}\n- Disparate impact — worst protected characteristic: {{output_payload.worst_di_ratio}} ({{output_payload.worst_di_characteristic}})\n- Equalized odds gap: {{output_payload.max_eod_gap}}\n- Annex IV documentation completeness: {{output_payload.annex_iv_completeness}}%\n- Days to August 2, 2026 deadline: {{output_payload.days_to_deadline}}\n- Open findings: {{output_payload.open_findings_count}}\n\nYour task: draft an AI Act conformity assessment memo (max 500 words). Structure: (1) High-risk classification basis (Annex III Part 5(b)), (2) Bias findings with Art. 10 data governance implications, (3) Technical documentation gaps vs Annex IV, (4) Remediation roadmap with deadline-weighted priorities, (5) Notified body readiness assessment.\n\nLabel all bias metrics as 'indicative, not production'. Conformity self-assessment is not a legal opinion — flag for qualified counsel review.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "attestation_mandate": {
      "display_name": "Attestation Mandate (GENIUS Act / Reserve)",
      "tool_ids": ["art-06-genius-act-reserve-attestation"],
      "regulatory_citations": [
        "GENIUS Act §4(a) — reserve composition: permitted asset classes",
        "GENIUS Act §4(b) — 1:1 reserve coverage requirement",
        "GENIUS Act §4(c) — monthly reserve report, CEO/CFO certification under AICPA 2025 Criteria",
        "GENIUS Act §4(d) — annual PCAOB audit for issuers > $50B outstanding",
        "FDIC NPRM April 2026 (GENIUS Act implementation)",
        "EU MiCA Art. 36–37 (e-money token reserve requirements — distinct from GENIUS Act)"
      ],
      "primary_tasks": ["regulatory_submission", "board_memo", "risk_memo", "remediation_plan"],
      "audiences": ["board", "risk_committee", "regulator", "ops"],
      "stochastic_framing_rules": [
        "Reserve coverage is a deterministic calculation given input figures — state exact coverage ratio, not a range",
        "Asset composition eligibility is rule-based — use ELIGIBLE/INELIGIBLE per GENIUS Act permitted-asset taxonomy",
        "CEO/CFO certification readiness is a pre-check only — actual certification requires qualified legal counsel",
        "PCAOB threshold is a binary test at $50B — state REQUIRED/NOT-REQUIRED with current outstanding amount",
        "Label all outputs as 'attestation readiness assessment' — not a substitute for PCAOB-audited reserve attestation"
      ],
      "escalation_conditions": [
        "Reserve coverage < 100% — immediate board escalation, regulatory notification obligation triggered",
        "Ineligible asset detected in reserve composition — divestiture required before next monthly report",
        "Outstanding > $50B and no PCAOB audit engaged — urgent audit engagement required",
        "CEO/CFO certification criteria not met — monthly report submission blocked",
        "FDIC NPRM interaction detected — legal review of deposit insurance implications required"
      ],
      "audience_formats": {
        "board": "Reserve coverage verdict, eligible vs. non-eligible asset breakdown, CEO/CFO certification readiness, GENIUS Act timeline",
        "risk_committee": "Reserve composition analysis, eligibility gap, remediation actions, PCAOB audit preparation status",
        "regulator": "Monthly reserve report in GENIUS Act / AICPA 2025 format, coverage computation, eligible asset certification",
        "ops": "Reserve rebalancing instructions, eligible asset list, reporting calendar, certification workflow"
      },
      "tone_modifiers": {
        "formal": "Lead with the attestation readiness determination. Reference GENIUS Act and AICPA 2025 Criteria.",
        "technical": "Include asset classification methodology, coverage computation, PCAOB preparation status.",
        "plain": "Explain whether reserves are sufficient and what needs to change before attestation."
      },
      "template_body": "You are a stablecoin compliance officer reviewing a ChainGraph GENIUS Act reserve attestation pre-check (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nReserve attestation summary:\n- Reserve coverage ratio: {{output_payload.reserve_coverage_ratio}} (requirement: 100%)\n- Coverage verdict: {{output_payload.coverage_verdict}}\n- Ineligible assets detected: {{output_payload.ineligible_asset_count}}\n- CEO/CFO certification readiness: {{output_payload.cert_readiness_verdict}}\n- PCAOB audit required: {{output_payload.pcaob_required}}\n- Days to next monthly report deadline: {{output_payload.days_to_report}}\n\nYour task: draft a monthly reserve attestation readiness memo (max 400 words). Structure: (1) Coverage compliance determination (cite GENIUS Act §4(a)–(b)), (2) Asset composition findings with ELIGIBLE/INELIGIBLE classifications, (3) CEO/CFO certification checklist status (AICPA 2025 Criteria), (4) PCAOB audit obligation, (5) Pre-submission remediation actions.\n\nThis is a pre-check only — label as 'attestation readiness assessment', not a substitute for PCAOB-audited reserve attestation.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "infrastructure_mandate": {
      "display_name": "Infrastructure Mandate (DORA)",
      "tool_ids": ["art-09-dora-incident-classifier"],
      "regulatory_citations": [
        "DORA (EU) 2022/2554 Art. 19 (major ICT incident classification criteria)",
        "DORA Art. 20 (initial notification 4h; intermediate 72h; final 1 month)",
        "DORA Art. 17 (ICT Risk Management Framework)",
        "DORA Art. 23 (classification of ICT-related incidents)",
        "DORA Art. 28 (critical third-party ICT providers — CTPP designation)",
        "EBA/GL/2023/07 (guidelines on ICT risk management under DORA)",
        "NIS2 Directive (EU 2022/2555) — cross-reference for significant incident thresholds",
        "FAPI 2.0 (Financial-grade API Security)"
      ],
      "primary_tasks": ["regulatory_submission", "risk_memo", "remediation_plan", "board_memo"],
      "audiences": ["board", "risk_committee", "regulator", "ops"],
      "stochastic_framing_rules": [
        "DORA incident classification is DETERMINISTIC — use MAJOR/SIGNIFICANT/BELOW-THRESHOLD only",
        "Reporting clock deadlines are deterministic calendar calculations — state exact datetime, not ranges",
        "Incident parameters are user inputs — classify based on inputs, do not independently estimate impact",
        "Do not qualify the classification with hedged language — regulators expect a clear determination"
      ],
      "escalation_conditions": [
        "Major incident classification — immediate CISO and board escalation, regulatory notification within 4 hours",
        "Cross-border component detected — notify all relevant competent authorities simultaneously",
        "CTPP involvement — additional DORA Art. 28 oversight obligations apply",
        "Transaction value > €100M affected — board-level notification alongside regulatory filing",
        "Downtime > 24h — activate full business continuity plan"
      ],
      "audience_formats": {
        "board": "Incident severity verdict, reporting obligation status, reputational risk assessment, board notification required?",
        "risk_committee": "Full DORA classification against all Art. 19 criteria, reporting timeline, ICT risk framework gap analysis",
        "regulator": "DORA Art. 19/20 notification in prescribed format, incident parameters, reporting clock status",
        "ops": "Reporting action list with deadlines, responsible contacts, notification drafts, containment protocol"
      },
      "tone_modifiers": {
        "formal": "Lead with the classification determination. Reference DORA articles directly.",
        "technical": "Detail each Art. 19 classification criterion and the incident parameter values.",
        "plain": "Explain what needs to be reported to whom and by when."
      },
      "template_body": "You are a DORA compliance officer reviewing a ChainGraph ICT incident classification (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nIncident classification:\n- DORA classification: {{output_payload.dora_classification}}\n- Reporting clock start: {{output_payload.clock_start}}\n- Initial notification deadline: {{output_payload.initial_deadline}} (4h from clock start, DORA Art. 20)\n- Intermediate report deadline: {{output_payload.intermediate_deadline}} (72h)\n- Final report deadline: {{output_payload.final_deadline}} (1 month)\n- Cross-border: {{output_payload.cross_border_flag}}\n- CTPP involved: {{output_payload.ctpp_flag}}\n\nYour task: draft a DORA initial notification memo (max 400 words). Structure: (1) Incident classification per DORA Art. 19 with criteria justification, (2) Reporting timeline with exact deadlines, (3) Competent authorities to notify (cite Art. 20), (4) Immediate containment actions.\n\nClassification is DETERMINISTIC — use MAJOR/SIGNIFICANT/BELOW-THRESHOLD only. State exact datetimes for deadlines.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "settlement_mandate": {
      "display_name": "Settlement Mandate (x402 / Rails)",
      "tool_ids": ["art-03-x402-settlement-modeler"],
      "regulatory_citations": [
        "x402 Protocol (HTTP 402 Payment Required — Coinbase CDP, 2025)",
        "CPMI-IOSCO Principles for Financial Market Infrastructures §8 (settlement finality)",
        "EU Instant Payments Regulation — 10-second settlement SLA (October 2025)",
        "EU Settlement Finality Directive (98/26/EC) — DLT application via DLT Pilot Regime",
        "ESMA DLT Pilot Regime (EU 2022/858) — on-chain settlement eligibility",
        "PSD2 Art. 77–80 (settlement timeframes for card/ACH/SWIFT)",
        "SWIFT CBPR+ ISO 20022 migration (2023–2025) — STP rate implications"
      ],
      "primary_tasks": ["risk_memo", "board_memo", "plain_english_summary", "period_comparison"],
      "audiences": ["board", "risk_committee", "quant", "ops"],
      "stochastic_framing_rules": [
        "Cost-per-transaction figures are model estimates based on published fee schedules — cite source date",
        "Settlement finality classifications are deterministic based on rail characteristics",
        "Cross-border compliance flags are heuristic — flag for legal review, not definitive compliance determination",
        "x402 active-agent estimates (~69k agents, 165M+ transactions) are market estimates as of mid-2026"
      ],
      "escalation_conditions": [
        "No rail achieves < 5% total cost for the requested corridor — escalate to treasury for bespoke FX negotiation",
        "All rails flag cross-border compliance concerns — legal review required before transaction execution",
        "Micropayment amount < $0.01 and only card/ACH eligible — transaction economically unviable",
        "x402 selected for transaction > $10,000 — Travel Rule threshold monitoring flag",
        "Finality gap > 60 seconds vs. IPR 10-second SLA"
      ],
      "audience_formats": {
        "board": "Rail recommendation and strategic rationale, cost/speed trade-off summary, key risk flags",
        "risk_committee": "Full cost and finality comparison across rails, x402 vs. card vs. bank analysis, FX risk assessment",
        "quant": "Quantitative model output, methodology notes, Monte Carlo parameters if applicable, sensitivity analysis",
        "ops": "Integration requirements per recommended rail, testing protocol, fallback rail configuration"
      },
      "tone_modifiers": {
        "formal": "Lead with the rail recommendation. Reference CPMI PFMIs and IPR requirements.",
        "technical": "Include finality mechanics, cost components, and x402 protocol parameters.",
        "plain": "Explain which payment rail to use and why, in practical terms."
      },
      "template_body": "You are a treasury operations analyst reviewing a ChainGraph settlement rail recommendation (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nRail recommendation:\n- Recommended rail: {{output_payload.recommended_rail}}\n- Estimated cost: {{output_payload.cost_estimate}}\n- Finality classification: {{output_payload.finality_classification}}\n- Cross-border flags: {{output_payload.cross_border_flags}}\n- x402 eligible: {{output_payload.x402_eligible}}\n\nYour task: write a treasury operations memo (max 300 words). Structure: (1) Rail recommendation with cost/finality rationale (cite CPMI-IOSCO PFMIs §8 and applicable directive), (2) Regulatory compliance flags, (3) Cross-border compliance checklist, (4) Implementation steps.\n\nCost estimates are model outputs based on published fee schedules — label as 'estimated, not contractual'. Finality classifications are deterministic.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "prompt_template": {
      "display_name": "Prompt Template (PTG-01)",
      "tool_ids": ["ptg-01-ap2-prompt-template-generator"],
      "regulatory_citations": [
        "EU AI Act Annex IV (technical documentation requirements)",
        "ISO/IEC 42001:2023 §6.1 (AI risk management documentation)",
        "NIST AI RMF Govern 1.1 (AI governance documentation)"
      ],
      "primary_tasks": ["plain_english_summary"],
      "audiences": ["board", "risk_committee", "regulator", "quant", "ops"],
      "stochastic_framing_rules": [
        "This is a bridge tool — no stochastic output framing applies",
        "ALWAYS cite the execution_hash from the source artifact in the prompt footer",
        "Distinguish between rule-based determinations (deterministic) and simulation results (stochastic) in system prompt framing"
      ],
      "escalation_conditions": [
        "Source artifact execution_hash cannot be verified — prompt the operator to re-run the source tool",
        "chain_depth > 3 — flag for human-in-the-loop review before narrative is submitted to regulator",
        "compliance_flags array is non-empty — escalate narrative review to qualified compliance officer"
      ],
      "audience_formats": {
        "board": "Summary of the prompt generated and the artifact it was derived from",
        "risk_committee": "Full prompt with provenance chain",
        "regulator": "Prompt with full chain of custody and execution hashes",
        "quant": "Technical breakdown of template registry entry applied",
        "ops": "Prompt text ready for use, deep link, copy instructions"
      },
      "tone_modifiers": {
        "formal": "Professional tone throughout.",
        "technical": "Include registry version and mandate_type classification details.",
        "plain": "Plain description of what was generated and how to use it."
      },
      "template_body": "You are a regulator-grade compliance narrative assistant operating under {{mandate_type}} framing. You have received a verified ChainGraph artifact with execution_hash {{execution_hash}} generated by {{tool_id}} v{{tool_version}} at {{generated_at}}.\n\nYour task: produce a concise, accurate compliance narrative from the attached artifact. Rules:\n1. Never hallucinate regulatory text. Cite only the regulatory references provided in the artifact's policy_parameters.\n2. Distinguish deterministic verdicts (rule-based) from probabilistic estimates (simulation-based).\n3. End every narrative with: 'This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload.'\n4. If compliance_flags is non-empty, list each flag as a numbered escalation item before the narrative body.\n5. Do not summarise input parameters — focus on the output_payload interpretation and its regulatory implications.\n\nArtifact: {{artifact_json}}"
    },
    "credit_assessment": {
      "display_name": "Credit Assessment (IRB / SA)",
      "tool_ids": [],
      "regulatory_citations": [
        "Basel F-IRB / A-IRB (Foundation / Advanced Internal Ratings-Based Approach)",
        "Basel SA-CR (Standardised Approach — Credit Risk)",
        "CRR3 / Basel 3.1 — credit risk (UK PRA go-live January 2027)",
        "EBA Guidelines on PD and LGD estimation"
      ],
      "primary_tasks": ["board_memo", "regulatory_submission", "risk_memo", "period_comparison"],
      "audiences": ["board", "risk_committee", "regulator", "quant"],
      "stochastic_framing_rules": [
        "PD/LGD/EAD are model estimates — label as 'model output, not regulatory determination'",
        "RWA computations are pre-validation — confirm with regulatory approval before use in capital reporting"
      ],
      "escalation_conditions": [
        "PD exceeds internal watch-list threshold",
        "LGD estimate above 45% (F-IRB floor for senior unsecured)",
        "RWA output floor binding under Basel 3.1",
        "Expected Loss > regulatory provision"
      ],
      "audience_formats": {
        "board": "Portfolio credit quality verdict, key risk concentrations, expected loss vs. provisions, strategic capital implications",
        "risk_committee": "Full PD/LGD/EAD/RWA breakdown, watch-list exposure, approach comparison, provisioning gap",
        "regulator": "IRB model output in prescribed format, backtesting summary, model validation status, regulatory approval status",
        "quant": "Full model output with methodology, parameter assumptions, convergence diagnostics, uncertainty ranges"
      },
      "tone_modifiers": {
        "formal": "Lead with the credit quality determination. Reference Basel approach and articles.",
        "technical": "Include PD/LGD/EAD parameters, RWA computation methodology, backtesting statistics.",
        "plain": "Explain credit quality in terms of likelihood of default and potential loss."
      },
      "template_body": "You are a credit risk analyst reviewing a ChainGraph credit assessment artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nCredit assessment summary:\n- Portfolio credit quality: {{output_payload.credit_quality_verdict}}\n- PD range: {{output_payload.pd_min}}% – {{output_payload.pd_max}}%\n- LGD estimate: {{output_payload.lgd_estimate}}%\n- RWA vs. regulatory floor: {{output_payload.rwa_vs_floor}}\n- Expected Loss vs. provisions: {{output_payload.el_vs_provisions}}\n\nYour task: write an ICAAP credit risk section (max 400 words). Structure: (1) Portfolio credit quality determination (cite Basel approach used), (2) PD/LGD/EAD parameter analysis, (3) Output floor binding assessment (Basel 3.1, deadline 2027-01-01), (4) Expected loss vs. provisioning gap, (5) Watch-list escalation flags.\n\nLabel all outputs as 'model output, not regulatory determination'.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "treasury_mandate": {
      "display_name": "Treasury Mandate",
      "tool_ids": [],
      "regulatory_citations": [
        "CPMI Principles for Financial Market Infrastructures",
        "FX settlement finality — CLS Group standards",
        "SEPA Credit Transfer Scheme Rulebook"
      ],
      "primary_tasks": ["board_memo", "risk_memo", "plain_english_summary"],
      "audiences": ["board", "risk_committee", "quant", "ops"],
      "stochastic_framing_rules": [
        "FX and rate estimates are model outputs — label as 'indicative'",
        "Settlement finality times are corridor estimates — actual finality depends on CLS/RTGS operating hours"
      ],
      "escalation_conditions": [
        "FX exposure above treasury limit",
        "Settlement finality > T+2 on any leg",
        "Liquidity buffer below minimum threshold"
      ],
      "audience_formats": {
        "board": "Treasury position summary, key risk exposures, FX and liquidity verdict",
        "risk_committee": "Full exposure breakdown, hedging gap analysis, settlement risk by corridor",
        "quant": "Quantitative model output, correlation assumptions, VaR contribution",
        "ops": "Settlement instructions, hedge execution checklist, corridor cutoff times"
      },
      "tone_modifiers": {
        "formal": "Lead with the treasury position determination.",
        "technical": "Include model parameters, hedge ratios, settlement mechanics.",
        "plain": "Explain the treasury position and what action is needed."
      },
      "template_body": "You are a treasury risk officer reviewing a ChainGraph treasury mandate artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nTreasury summary:\n- Position verdict: {{output_payload.position_verdict}}\n- FX exposure: {{output_payload.fx_exposure}}\n- Settlement finality: {{output_payload.settlement_finality}}\n- Liquidity buffer: {{output_payload.liquidity_buffer}}\n\nYour task: write a treasury risk memo (max 300 words). Structure: (1) Position determination (cite CPMI PFMIs), (2) FX exposure and hedging gap, (3) Settlement finality assessment, (4) Required actions with urgency flags.\n\nLabel FX and rate estimates as 'indicative'. Settlement finality times are corridor estimates.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    },
    "account_mandate": {
      "display_name": "Account Mandate",
      "tool_ids": [],
      "regulatory_citations": [
        "GDPR Article 25 (Privacy by Design and by Default)",
        "SEPA Direct Debit Scheme Rulebook",
        "PSD2 Account Information Services provisions"
      ],
      "primary_tasks": ["risk_memo", "remediation_plan", "plain_english_summary"],
      "audiences": ["risk_committee", "regulator", "ops"],
      "stochastic_framing_rules": [
        "Account data is synthetic — label all account references as test data"
      ],
      "escalation_conditions": [
        "GDPR Art. 25 privacy-by-design requirement not met",
        "SDD mandate validity expiry",
        "Account information access scope exceeds PSD2 authorization"
      ],
      "audience_formats": {
        "risk_committee": "Account risk assessment, GDPR compliance status, mandate validity",
        "regulator": "PSD2 / GDPR compliance statement, access scope certification",
        "ops": "Mandate renewal checklist, GDPR controls verification, SDD operational protocol"
      },
      "tone_modifiers": {
        "formal": "Lead with the account compliance determination.",
        "technical": "Reference GDPR articles and PSD2 provisions explicitly.",
        "plain": "Explain the account status and what needs to be done."
      },
      "template_body": "You are a compliance officer reviewing a ChainGraph account mandate artifact (execution_hash: {{execution_hash}}, tool: {{tool_id}}).\n\nAccount mandate summary:\n- Compliance verdict: {{output_payload.compliance_verdict}}\n- GDPR Art. 25 status: {{output_payload.gdpr_status}}\n- SDD mandate validity: {{output_payload.sdd_validity}}\n- PSD2 access scope: {{output_payload.psd2_scope_verdict}}\n\nYour task: write a compliance memo (max 300 words). Structure: (1) Account compliance determination (cite GDPR Art. 25 and PSD2 provisions), (2) SDD mandate status, (3) Remediation actions required.\n\nAll account data is synthetic — label all references as test data.\n\nAudit anchor: This analysis is anchored to execution_hash {{execution_hash}}. To verify, re-run {{tool_id}} with the same policy_parameters and confirm the SHA-256 hash over sorted-key JSON of policy_parameters + output_payload."
    }
  }
}