Agentic AI Testing / Quiz

Context Engineering for Tool-Heavy Agents

Every tool result is more tokens. Left unmanaged, an agent poisons its own context and rots mid-task. Engineer the window, don't just fill it.

Difficulty
Easy
Format
Quiz
Points
100
Estimate
9 min

// MISSION BRIEF

Your Mission

A tool-using agent accumulates context fast: each tool call returns data that stacks up in the window. Poor context engineering, dumping raw results, keeping stale data, leaking one sub-agent's junk into another, degrades the agent step by step. This quiz covers managing the window deliberately.

Scoring: equal weight per question.

// FIRST CONTACT

Battle teaser

A tool returns a 50,000-token API dump every call, and the agent degrades after a few steps. The fix:

  1. AAdd more tools
  2. BIncrease temperature
  3. CLengthen the system prompt
  4. DSummarize/filter tool results to the relevant fields before adding them to context, rather than pasting raw payloads
Answers, scoring, hints, and the full battle stay sealed.

// SKILL TAGS

agenticcontext-engineeringmemorymulti-agent