STRANDS GRAPH RUNTIMENO CHAIN-OF-THOUGHT LEAK

Live Agent Graph Execution

Displaying real-time execution telemetry, deterministic policy evaluations, tool invocations, and confidence scoring.

Execution Pipeline10 Nodes Evaluated • 0 Infinite Loops

Intake Agent

report_get

Category: streetlights, asset identified

0.8s97% conf

Location & Vision

location_reverse_geocode

North Gate 2, University Road

1.1s94% conf

Duplicate Detection

report_find_nearby

Distance: 24.5m (Auto-merged into RG-M-201)

0.6s96% conf

Physical Verification

evidence_get_metadata

Physical asset existence verified

1.2s96% conf

Deterministic Risk Engine

policy_lookup

Score: 85 (HIGH risk band)

0.3s95% conf

Department Ownership

department_lookup

Dept: electrical, Skill: street_lighting

0.5s98% conf

Mission Dispatch

mission_create

Mission RG-M-201 created with 8h SLA

0.4s100% conf

Worker Match Engine

worker_get_available

Matched Ahmed Khan (Score: 0.96)

0.5s96% conf

Guardian Policy Check

guardian_evaluate

Action permitted: routine low-risk dispatch

0.2s100% conf

Proof-of-Repair Verifier

evidence_compare_vision

Autonomous closure recommended

1.4s96% conf

NODE INSPECTOR

Deterministic Risk Engine

COMPLETED

Agent ClassRiskAgent

Invoked Toolpolicy_lookup()

Execution Latency0.3s

Confidence Rating95.0%

Structured Outcome

Score: 85 (HIGH risk band)

Privacy Safety Rule: Private LLM chain-of-thought is excluded from logs and client state. Only deterministic operational metadata is exposed.

AgentCore Runtime • us-east-1 • Strands Graph Engine