◈ AWS AGENTCORE BLUEPRINTS
PART V · MULTI-AGENT AND ARCHITECTURE · CHAPTER 33 / 36
TOCCONSULTING.FR
AMAZON BEDROCK / PART V, MULTI-AGENT AND ARCHITECTURE

SECURITY AND THREAT MODEL

CHAPTER 33 / 36
IN ONE LINE the OWASP risks, mapped to AgentCore controls
REV 2026.07

An agent is riskier than a chatbot because it acts: it sends email, queries databases, calls APIs, spends money. So the things that can go wrong are bigger. This chapter takes the standard list of LLM risks (the OWASP Top 10 for LLM Applications, 2025) and maps each one to the AgentCore control that blunts it, and to the part that stays your job.

New to these words? Read chapter 02 "Key terms". Related: Identity (chapter 13), Policy (chapter 23), Common mistakes (chapter 03).
[ 1 ] THE OWASP LLM TOP 10 (2025), MAPPED TO AGENTCORE
RISK (OWASP 2025)IN PLAIN WORDSAGENTCORE CONTROLSTILL YOUR JOB
LLM01 Prompt injectionHidden instructions in a message or web page hijack the agent.Bedrock Guardrails filter inputs; Policy blocks disallowed tool calls; least-privilege scopes limit the damage.Add Guardrails, scope tools tightly, validate external content.
LLM02 Sensitive info disclosureThe agent leaks secrets or personal data.KMS encryption; Identity token vault (the agent never holds raw secrets); Guardrails PII filters.Classify data, enable PII filters, keep secrets out of Memory metadata.
LLM03 Supply chainA poisoned model, tool, or dependency gets in.Registry (only vetted, approved tools and MCP servers are discoverable); IAM on who can publish.Vet dependencies; use the Registry approval workflow.
LLM04 Data / model poisoningBad data corrupts the agent's memory or behavior.Validate inputs at CreateEvent; Guardrails; Memory namespace isolation.Check what enters Memory; guard untrusted inputs.
LLM05 Improper output handlingThe agent's output is used unsafely downstream (SQL, code).Policy / Cedar gates every tool call; Code Interpreter sandbox; the gateway SQL-injection guard pattern.Treat output as untrusted; parameterize queries.
LLM06 Excessive agencyThe agent is simply allowed to do too much.Policy (allow/deny per tool and parameter); least-privilege scopes; human in the loop (Browser live view).Grant least privilege; add approval gates for risky actions.
LLM07 System prompt leakageSecrets hidden in the system prompt get exposed.Keep secrets out of the prompt, use the Identity vault; Guardrails.Never put secrets, or rules-as-secrets, in the prompt.
LLM08 Vector / embedding weaknessesMemory or retrieval is manipulated, or leaks across tenants.Memory namespace isolation and access control; per-tenant scoping (see chapter 32).Scope retrieval to the right namespace and tenant.
LLM09 MisinformationThe agent confidently makes things up.Evaluations (LLM-as-a-judge scoring); Observability traces; grounding via tools and Memory.Evaluate every version, ground answers, monitor quality.
LLM10 Unbounded consumptionRunaway cost, or denial of service.Runtime session limits (15-min idle, 8-h max); Payments spending limits; Observability cost alarms.Set budgets, quotas, and CloudWatch alarms.
Across all of these, the per-session microVM limits the blast radius of any single compromise, and CloudTrail + Observability give you the trail to detect and investigate.
[ 2 ] DEFENSE IN DEPTH, THE LAYERS A REQUEST PASSES THROUGH
AUDIT

CloudTrail, Observability traces. See and prove what happened.

NETWORK

VPC, PrivateLink. Keep traffic off the public internet.

IDENTITY

Workload identity, token vault, IAM least privilege.

ISOLATION

Per-session microVM, wiped at the end. No cross-session bleed.

GOVERNANCE

Policy / Cedar on tool calls; Bedrock Guardrails on model text.

DATA

TLS 1.2+ in transit; KMS at rest; customer-managed keys for Memory, Gateway, and the Identity vault.

[ 3 ] WHO SECURES WHAT (SHARED RESPONSIBILITY)
AWS SECURES of the cloud
The infrastructure, the microVM isolation, patching, and the managed services themselves, tested under AWS compliance programs.
YOU SECURE in the cloud
Your prompts, your tool scopes, your Guardrails and Policy config, data classification, output handling, and budgets.
Plain definitions: chapter 02. Controls in detail: Identity (chapter 13), Policy (chapter 23), Observability (chapter 19), Evaluations (chapter 21).
SRC: OWASP Top 10 for LLM Applications 2025 (genai.owasp.org) + AWS AgentCore security docs + awslabs samples. Mapping is guidance. DRAWN 2026.06, REV 2026.07.26
◄ PREVIOUS · CH 32
Multi-Tenant Agents, Plain Language
NEXT · CH 34 ►
Governance at Org Scale, Plain Language
AWS AGENTCORE BLUEPRINTS · CONTENTS
Back to tocconsulting.fr Cover & Table of Contents
PART I · FOUNDATIONS
01AgentCore Overview, Plain LanguageCONCEPT02AgentCore, Key Terms in Plain LanguageCONCEPT03AgentCore, Common Mistakes ExplainedCONCEPT04Inside an AgentCore Agent, BlueprintCONCEPT05Harness, Practical 1, Strands PathPRACTICAL06Harness, Practical 2, Managed HarnessPRACTICAL
PART II · CORE SERVICES
07AgentCore Runtime, Plain LanguageCONCEPT08AgentCore Runtime, PracticalPRACTICAL09AgentCore Memory, Plain LanguageCONCEPT10AgentCore Memory, PracticalPRACTICAL11AgentCore Gateway, Plain LanguageCONCEPT12AgentCore Gateway, PracticalPRACTICAL13AgentCore Identity, Plain LanguageCONCEPT14AgentCore Identity, PracticalPRACTICAL
PART III · TOOLS AND OPERATIONS
15AgentCore Code Interpreter, Plain LanguageCONCEPT16AgentCore Code Interpreter, PracticalPRACTICAL17AgentCore Browser, Plain LanguageCONCEPT18AgentCore Browser, PracticalPRACTICAL19AgentCore Observability, Plain LanguageCONCEPT20AgentCore Observability, PracticalPRACTICAL21AgentCore Evaluations, Plain LanguageCONCEPT22AgentCore Evaluations, PracticalPRACTICAL
PART IV · GOVERNANCE AND TRANSACTIONS
23AgentCore Policy, Plain LanguageCONCEPT24AgentCore Policy, PracticalPRACTICAL25AgentCore Registry, Plain LanguageCONCEPT26AgentCore Registry, PracticalPRACTICAL27AgentCore Payments, Plain LanguageCONCEPT28AgentCore Payments, PracticalPRACTICAL
PART V · MULTI-AGENT AND ARCHITECTURE
29A2A and Multi-Agent, Plain LanguageCONCEPT30A2A and Multi-Agent, PracticalPRACTICAL31Architecture Patterns, Plain LanguageCONCEPT32Multi-Tenant Agents, Plain LanguageCONCEPT33Agent Security and Threat Model, Plain LanguageCONCEPT34Governance at Org Scale, Plain LanguageCONCEPT35Governance at Org Scale, PracticalPRACTICAL36Multi-Region and Distribution, Plain LanguageCONCEPT