Agentic AI: Write Tool Descriptions That Agents Use Correctly

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-agentic-tool-discoverability-for-agents-r8a1-creators

1. A developer is naming tools for an agent system. Which naming approach best matches the recommendation from the material on tool discoverability?

   1. Use descriptive nouns that capture what the tool does, like 'WeatherChecker'
   2. Use verb-object patterns that specify the action, like 'get_weather'
   3. Use acronyms that are easy to type quickly, like 'WTHR
   4. Use generic names that work for any function, like 'Tool1'

2. What should a complete tool description explicitly address to help an agent decide when to invoke that tool?

   1. Only what the tool does when it works correctly
   2. When to use the tool AND when NOT to use the tool
   3. The code implementation details of the tool
   4. The developer's personal notes about the tool

3. A developer documents a parameter as 'limit: number'. Why is this insufficient according to the principles of effective parameter documentation?

   1. Parameters should never be documented
   2. The documentation should include constraints, units, and example values
   3. The parameter name should be longer
   4. Only return parameters need documentation

4. What does the material identify as a key limitation when using AI to improve tool descriptions?

   1. AI cannot write descriptions that are detailed enough
   2. AI cannot test whether the rewrite actually improves agent behavior
   3. AI will always make descriptions too long
   4. AI cannot identify which tools exist in the system

5. What is tool overlap, and what should you do when you identify it according to the material?

   1. Overlap means tools have the same name; rename one to be unique
   2. Overlap means multiple tools could solve the same task; flag it for human review
   3. Overlap means tools share code; merge them into one tool
   4. Overlap means tools are unused; delete the extra tools

6. Why is including both a good and a bad call example in tool documentation considered valuable?

   1. It makes the documentation look more complete
   2. Good examples show ideal usage while bad examples illustrate common mistakes agents might make
   3. The examples are only for human developers to read
   4. Bad examples confuse agents and should be omitted

7. What does the material identify as a bias that can affect tool selection even with good descriptions?

   1. Agents prefer tools with longer descriptions
   2. Models are biased toward whichever tool is listed first in the tool list
   3. Agents always choose the tool with the most parameters
   4. Models cannot read descriptions longer than 100 words

8. A developer claims their tool description rewrite improved the agent's performance. What evidence would actually prove this claim according to the material?

   1. The developer feels confident the agent will work better
   2. A comparison of tool-choice accuracy before and after the rewrite
   3. The new description is longer than the old one
   4. Other developers said the new description looks better

9. The material mentions that cleaner tool docs can outperform model upgrades. What does this imply about tool descriptions?

   1. Tool descriptions are more important than the underlying model architecture
   2. Well-written descriptions can compensate for using older or simpler models
   3. Model upgrades never help agent performance
   4. Tool descriptions should be changed every time a new model is released

10. What should a known-failure-mode note in a tool description address?

    1. How the tool was programmed incorrectly
    2. Common ways the tool is likely to be misused or produce errors
    3. The developer's email for reporting bugs
    4. Historical versions of the tool that no longer work

11. What is the relationship between tool description quality and agent performance according to the material's core premise?

    1. Most agent tool-misuse comes from sloppy tool descriptions
    2. Most agent tool-misuse comes from choosing too many tools
    3. Most agent tool-misuse comes from using the wrong model
    4. Most agent tool-misuse comes from network errors

12. When documenting a parameter, what three elements should be included according to the material?

    1. Name, type, and default value
    2. Constraints, units, and examples
    3. Required flag, help text, and version
    4. Data type, programming language, and API endpoint

13. A developer pastes their current tool definitions and asks an AI to rewrite them. What specific additional requests should they include to follow the material's approach?

    1. Just ask for cleaner wording
    2. Request when-not-to-use sections, parameter examples, and known-failure-mode notes
    3. Ask for shorter descriptions to save tokens
    4. Request the AI to test if the new descriptions work

14. What does the material say about wishful thinking in tool description improvement?

    1. It's fine to hope descriptions work better after rewriting
    2. Without measurement, you are wishful-thinking your prompt
    3. Descriptions should focus on hopeful language
    4. Wishful thinking improves agent confidence

15. What distinguishes an effective tool description from an ineffective one according to the material?

    1. Effective descriptions are longer and more detailed
    2. Effective descriptions are unambiguous and actionable for the agent
    3. Effective descriptions use technical jargon
    4. Effective descriptions include the tool's source code