Deterministic 3-Step Workflow

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R24 — LangGraph persistence, checkpointing & interrupts

   R41 — Agentic workflow risks

   R09 — PostgreSQL transaction isolation

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

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Implement P3's deterministic workflow (extract facts -> parallel generation -> score/select) separate from free-form agent loops. Define strict state-transition invariants. *Lab:* Test invalid state transitions and state-replay attempts.

   Add LangGraph (or equivalent) durable checkpointing. Cryptographically bind every checkpoint to user_id, tenant_id, workflow_id, and auth_scopes. *Lab:* Attempt cross-user checkpoint restoration and prove it fails closed.

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

3. 03

   Security lab

   60 min· attack your own systemtimer

   No separate task text for this block on day 34 — the sourced work sits in the block above. Keep the time box.

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

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Workflow state-invariant tests + cross-tenant checkpoint isolation tests.

   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

• R09PostgreSQL Documentation↗ postgresql.org

  Foreign-key constraints, indexes, ACID transactions, row-level security (RLS), and query planning.

• R24LangGraph Documentation↗ docs.langchain.com

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

• R41OWASP Agentic AI Resources↗ genai.owasp.org

  Agentic threat taxonomy: tool misuse, excessive agency, memory poisoning, inter-agent trust, and human oversight.

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

• docsLangChain · ~25 min read
  Checkpointers - Docs by LangChain

  Documents thread-scoped checkpoints, the three durability modes and `EncryptedSerializer`, which are the primitives for binding a checkpoint record to a thread plus its owning identity.

  R24checked 2026-11-11
• videoLangChain · ~12 min
  LangGraph: Persistence

  Shows a checkpointer writing graph state at every super-step and resuming by thread ID, the mechanism you must gate on user_id and tenant_id before restore.

  R24checked 2026-11-11
• videoRenjith KN · ~15 min
  PostgreSQL Row-Level Security for Multi-Tenancy

  Demonstrates enabling RLS and writing a policy against a session tenant variable, the database-level backstop when a checkpoint restore is attempted across tenants.

  R09checked 2026-11-11
• articleCrunchy Data · 3 min read
  Row Level Security for Tenants in Postgres

  Shortest complete walkthrough of ENABLE ROW LEVEL SECURITY plus a tenant policy keyed to a per-request session variable, so replay of another tenant's checkpoint returns zero rows.

  R09checked 2026-11-11
• labCrunchy Data · ~20 min browser lab
  Row Level Security | Tutorials | Crunchy Data

  Interactive single-node Postgres where you SET ROLE and watch policies filter rows, letting you reproduce the cross-tenant restore attempt and see it fail closed.

  R09checked 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.