MISSION 01 · Agent Architecture
Agent, Tool Calling & Security Architecture
Agent→tool selection→schema→args validation→Backend API architecture and AI is not superuser
NOT STARTEDOpen
Week 7 operating page
Agent + HITL
Khóa tool authentication, permission boundaries, pending-only writes, HITL, audit và red-team.
AVAILABLE TO READ · LOCKED FOR PASS
Đọc và bookmark được phép; mọi mutation vẫn bị khóa server-side cho đến khi week trước PASS.
Missions
0/7
Weekly quiz
0%
PASS ≥ 80%
Evidence
0
learner artifacts
Missions
MISSION 02 · Agent Tools
Read Tools: Schema, Limits, Auth & Audit
Read tool such as get_low_stock_products, argument schema/limits, authorization, row limit and audit
NOT STARTEDOpen
MISSION 03 · Agent Write Security
Write Tools: Pending-only State & HITL
create_purchase_request tool, source=AI, DRAFT/PENDING, no approve tool, Manager approval and workflow continuation
NOT STARTEDOpen
MISSION 04 · Agent Security
Agent Service Authentication, RBAC & Permission Boundary
Agent service auth, user actor context, Backend RBAC, tool permission, expired token and 401/403 tests
NOT STARTEDOpen
MISSION 05 · Agent Operations
Agent Audit, Limits, Timeout & Resilience
Agent audit actor/tool/sanitized args/result/timestamp/correlation/latency, rate/row/quantity limits, timeout/malformed output and retry safety
NOT STARTEDOpen
MISSION 06 · Agent Red-team
Agent Red-team: Injection, Privilege & Secret Defenses
Ignore-rules, self-approve, secret extraction, SQL, huge quantity, malicious retrieved text, wrong role, expired token, timeout and malformed output red-team
NOT STARTEDOpen
MISSION 07 · Agent QA
Agent E2E, Security Closure & Configuration Freeze Prep
Two controlled tools, E2E Agent workflow, DRAFT/PENDING/Manager approval, AI audit, red-team report and final config inventory prep
NOT STARTEDOpen