OpenChainGraph · Agentic Commerce · Chain

Agentic Bond-Trading Mandate Receipt

Three-step chain: compile a work mandate (size / spread / issuer limits), simulate the spend policy against it, then run the agentic mandate sandbox against synthetic bond-trade transactions. Any breach halts the run and opens an escalation record; a clean run emits the per-session mandate receipt.

Linear · 3 steps §22.8 Escalation Gate Hash-Anchored §4 OCG v0.8.7 Client-Side · Zero PII
Who this is for: LTX / BondGPT product and heads of AI governance at dealer e-trading desks who need a replayable receipt that an agentic bond-trading mandate ran within its declared spend-policy bounds before it touches a live order.
Honest Proof-Scope Legend

Each step's receipt entry carries that kernel's existing proof status honestly — this chain composes only already-shipped, already-proven kernels; there is no new groth16 proving step for this composition. The chain's claim_strength is the minimum across steps: it is replayable-by-hash, not a fresh cryptographic proof of the composition itself.

§22.8 Escalation EMIT/HALT

When the sandbox gate fires escalate (any policy breach), the run halts immediately and an escalation record opens with a deterministic record_hash. Closure of that record is handled by verify_escalation_closure on the anchor worker — this chain references that tail, it does not reimplement it. A clean sandbox run (no breach) emits the per-session bond-trading mandate receipt: mandate bounds, sandbox verdict, gate decision, and escalation ref if any.

Computing…
Pending
DAG · 3 steps · art-274 → art-02 → rbe-06 (gate: breach → escalate)
STEP 1 Work Mandate Compiler
Compiles size, spread, and issuer limits into a structured work mandate for the trading session.
→ Handoff: mandate bounds feed Stage 2 spend-policy simulation
🔗 art-274-compile-work-mandate
MCP Call · compile_work_mandate
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "compile_work_mandate",
    "arguments": {
      "session_id": "SYNTHETIC-SESSION-001",
      "max_order_size_usd": 5000000,
      "max_spread_bps": 25,
      "issuer_limits": [{"issuer":"SYNTHETIC-ISSUER-A","max_notional_usd":2000000}]
    }
  },
  "id": 1
}
STEP 2 Agent Spend-Policy Simulator
Runs the compiled mandate against synthetic bond-trade transactions to produce a spend-policy verdict before the sandbox stage.
→ Handoff: spend_policy_verdict and synthetic-transaction results feed Stage 3 sandbox run
🔗 art-02-agent-spend-policy-simulator
MCP Call · simulate_spend_policy
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "simulate_spend_policy",
    "arguments": {
      "mandate_ref": "SYNTHETIC-SESSION-001",
      "synthetic_transactions": [{"notional_usd":1500000,"spread_bps":18,"issuer":"SYNTHETIC-ISSUER-A"}]
    }
  },
  "id": 2
}
STEP 3 Agentic Mandate Sandbox
Runs the mandate + spend-policy bounds through the agentic mandate sandbox against synthetic transactions. Terminal gate stage.
🏁 Terminal stage · GATE: breach → escalate (HALT + open record); default → emit receipt
§22.8.1 Decision Gate
rule[0]: if output.breach == true  →  next: escalate   (§22.8.2 HALT + open record)
default:                            →  next: receipt    (emit bond-trading mandate receipt)
Default emits: bond-trading mandate receipt
🔗 rbe-06-agentic-mandate-sandbox
MCP Call · agentic_mandate_sandbox
{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "agentic_mandate_sandbox",
    "arguments": {
      "mandate_ref": "SYNTHETIC-SESSION-001",
      "spend_policy_verdict_ref": "step-2-output"
    }
  },
  "id": 3
}
Export Chain Artifact
Download a §4-compliant chain artifact with execution hash, step definitions, and gate decision. Hash is deterministic over canonical {policy_parameters, output_payload} per RFC 8785/JCS.