Implement token counting

1. 01

   Concepts & system design

   60 min· read · map the mechanismtimer

   R04 — OWASP LLM10: Unbounded Consumption

   R03 — OWASP API4: Unrestricted Resource Consumption

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

2. 02

   Project implementation

   120 min· build · integrate · debugtimer

   Implement strict input token counting, history truncation, max_tokens output caps, and per-request context budgeting.

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

3. 03

   Security lab

   60 min· attack your own systemtimer

   Study denial-of-wallet and cost amplification; test oversized inputs and rapid context-stuffing requests.

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

4. 04

   Verification & evidence

   30 min· test · record evidencetimer

   Per-request token budget design and rejection tests for oversized payloads.

   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

• R03OWASP API Security Top 10↗ owasp.org

  BOLA (API1), Broken Auth (API2), Unrestricted Resource Consumption (API4), SSRF (API7), Misconfiguration (API8).

• R04OWASP Top 10 for LLM Apps↗ genai.owasp.org

  Prompt injection, sensitive info disclosure, improper output handling, excessive agency, vector weaknesses, unbounded consumption.

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

• docsOWASP API Security Top 10
  API4:2023 Unrestricted Resource Consumption

  Frames missing max_tokens, missing history caps, and per-request cost as an API-level denial-of-service and economic abuse risk, with the control set you implement as a context budget.

  R03M5moved 2026-11-11
• videoIBM Technology · 11:31
  What is a Context Window? Unlocking LLM Secrets

  Explains the token window as a fixed budget shared by system prompt, history, retrieved content, and reserved output, which is the arithmetic behind choosing a truncation point.

  M5checked 2026-11-11
• videoAI Engineer · 16:17
  How we solved Context Management in Agents — Sally-Ann Delucia

  Describes production strategies for keeping history inside the window (truncate, summarize, offload) and the failure modes of each, which is the tradeoff your truncation policy encodes.

  M5S7checked 2026-11-11
• articleLangChain Blog
  Context Engineering

  Taxonomizes context strategies (write, select, compress, isolate) and explains compaction and summarization, giving you named options instead of an ad hoc character cut.

  M5moved 2026-11-11
• labOpenAI Cookbook
  How to count tokens with Tiktoken

  Runnable code that counts message-array tokens including per-message overhead, which is the measurement you need before enforcing a max input budget.

  M5R03moved 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.