Run the complete Project 3 agent

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R21 — MCP

   R22 — OAuth

   R24 — LangGraph

   R33 — Playwright

   Read the assigned sections; map the mechanism, threat model and trade-offs.

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Clone/run Project 3 from a clean environment following only your README instructions.

   Build, integrate, debug and improve the increment in the active project.

3. 03

   Security lab

   60 min· attack your own systemtimer

   Execute the complete P3 acceptance suite: deterministic workflow & bounded agent tests, D2 Sandboxed MCP Proxy suite, D3 Inter-Agent Sanitization Bus suite, D7 Delegated Agent OAuth Broker & TOCTOU approval tests, D9 Computer-Use Pixel Auditor & C6 human-takeover suite, C4 sandboxed code-diagnosis workflow, D11 Cryptographic Trajectory Ledger chain verification, M10 misuse suite, D8 PyRIT CI gate (C12), and D12 SBOM scan.

   Test the abuse case, failure path, authorization bypass or resource limit.

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Clean-room build log, passing security/trajectory test report, and verified README for P3.

   Add or run regression tests; record the 9-point evidence record.

5. 05

   Deliberate practice & Socratic review

   30 min· explain aloud · update gapstimer

   Answer the 8 Socratic questions without notes: What? Why? How? What if? Why not? How can I prove it? Where can it break? What have I learned?

   Re-explain the hardest concept aloud, inspect one edge case, update the gap log.

6. total5 h

• R21Model Context Protocol (MCP)↗ modelcontextprotocol.io

  Official specification: client/server lifecycle, JSON-RPC messages, tool discovery, authorization, and trust boundaries.

• R22OAuth 2.0 Security BCP (RFC 9700)↗ rfc-editor.org

  RFC 6749 core flows + RFC 9700 security best practices: narrow scopes, audience restriction, token leakage, and replay.

• R24LangGraph Documentation↗ docs.langchain.com

  State graphs, durable checkpointing, human-in-the-loop interrupts, retries, and resumability.

• R33Playwright Documentation↗ playwright.dev

  Isolated BrowserContext, downloads, permissions, screenshots, bounding boxes, locators, and tracing.

Read only the named sections needed for today's task. Reading sits inside the 60-minute concepts block; it is not extra time.

• docsModel Context Protocol · Specification
  Model Context Protocol Specification

  The normative client/server lifecycle, JSON-RPC message shapes, tool discovery, and authorization boundaries — the trust boundary D2's sandboxed proxy has to enforce rather than assume.

  R21moved 2026-11-11
• videoMarco Codes · 13 min
  Model Context Protocol (MCP) - Explained

  Traces one full MCP transaction across host, client, and server with the tools/resources/prompts primitives, which is the concrete picture of what crosses the boundary on each tool call.

  R21checked 2026-11-11
• videoLangChain · Breakpoint demo
  LangGraph Agents - Human-In-The-Loop Breakpoints

  Shows interrupt_before breakpoints that pause a graph before a sensitive node and resume after approval, the mechanism behind D7's consent step and P3's human-oversight gate.

  R24checked 2026-11-11
• docsRFC Editor / IETF · ~60 pages
  RFC 9700: Best Current Practice for OAuth 2.0 Security

  The normative BCP for PKCE, exact redirect URI matching, audience restriction, sender-constrained tokens, and replay defence — the requirement set D7's delegated broker is judged against.

  R22moved 2026-11-11
• docsLangChain · Concept + API docs
  Checkpointers

  Explains per-super-step checkpoints, thread IDs, and the exit/async/sync durability modes, which determine whether a crashed agent resumes from state or replays side effects.

  R24checked 2026-11-11

A lab is complete only when it produces evidence. Installing PyRIT, Garak, Presidio or TransformerLens is not completion. Saved on this device only.