Week 16 Master Guide
Week 16: Agentic Guardrails, Code Sandboxing, Observability & Month 4 Capstone
Week 16 Daily Breakdown (5 Days)
Day 1
Day 1: Code Execution Sandboxes, AST Validation & Escape Prevention
1 code file(s) • 1 Q&ADay 2
Day 2: Policy-as-Code Guardrails (Input, Output & Tool Rails)
1 code file(s) • 1 Q&ADay 3
Day 3: OpenTelemetry Agent Tracing, Hierarchical Spans & Audit Logs
1 code file(s) • 1 Q&ADay 4
Day 4: Chaos Engineering for Agents (Tool Outages, Rate Limits & Fallbacks)
1 code file(s) • 1 Q&ADay 5
Days 5–7: Month 4 Capstone Project — Autonomous Agentic MCP Server & Infrastructure Copilot
1 code file(s) • 0 Q&AWeek 16: Agentic Guardrails, Code Sandboxing, Observability & Month 4 Capstone
Weekly Learning Objectives
- Day 1 (
day01_code_execution_sandboxes_ast_validation_and_resource_limits): Static Python AST Sandbox Validation blocking dangerous imports (os,subprocess,socket), dunder attribute traversal (__subclasses__,__globals__), and arbitrary builtins (eval,exec,open). - Day 2 (
day02_policy_as_code_guardrails_input_output_and_tool_rails): Deterministic Policy-as-Code Tool Guardrail Engine enforcing RBAC, environment blast-radius rules, and argument constraints before any MCP tool executes. - Day 3 (
day03_opentelemetry_agent_tracing_spans_and_audit_logs): OpenTelemetry-compliant Hierarchical Span Tracer (trace_id,span_id,parent_span_id, latency, token/cost attribution, and tamper-evident audit logs). - Day 4 (
day04_chaos_engineering_for_agents_tool_outages_and_fallbacks): Agent Chaos Monkey & Fallback Recovery Tester simulating MCP server timeouts, rate limits (429), and corrupted JSON payloads. - Days 5–7 (
day05_to_07_month4_capstone_agentic_mcp_server&capstones/month04_agentic_mcp_server): Production Month 4 Capstone Project — Autonomous Agentic MCP Server & Infrastructure Copilot uniting Weeks 13, 14, 15, and 16.