Prompt Harness

Don't Show AI What It Doesn't Need

If you have 100 tools and cram all their descriptions into the System Prompt, just the tool descriptions alone can consume tens of thousands of tokens. AI, like humans, loses focus when given too much information.

Full Load vs. Lazy Load

Full Load

~34,000 tk

Lazy Load

~2,800 tk

Drag to See the Difference

Adjust tool count to see how Token cost changes

Number of tools: 50

17,500

Full Load Tokens

1,550

Lazy Load Tokens

91%

Savings

How to Implement Lazy Loading

Three steps:

① First turn: names only
The System Prompt only includes tool names + a one-line summary. The AI just needs to know the capability exists.

② Expand on demand
When the AI decides to call a tool, the system dynamically injects the full description and parameter schema.

③ Retract after use
After the tool is used, the next turn no longer includes the full description — back to names only.

Analogy: It's like a company directory — you don't need everyone's full resume, just their name and role. When you actually need to work with someone, you look up the details.

AI, like humans, loses focus when given too much information. Give what's needed, expand on demand. Lazy loading tool descriptions both saves tokens and improves accuracy.