Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →saffron-health/prompting
Guide for writing effective system prompts for LLM agents.
Passed 16 of 22 test cases with the skill, and 5 of 22 without it.
and 15 more in the full report
11 fixed5 passed either way6 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
Write a system prompt for an incident response bot that monitors server alerts and searches log files upon request.
| Checked | Without skill | With skill |
|---|---|---|
| Uses bracketed action descriptions in examples. | Missed | Passed |
| Excludes italic styling. | Missed | Passed |
The instructions your model reads when the skill loads.
LLMs are intelligent by default. System prompts set direction and impose constraints, not explain reasoning.
Start minimal. Observe failures. Add targeted fixes. Every instruction must justify its token cost by solving a real problem.
Do not explain existing capabilities, list obvious practices, add preventive instructions, or repeat information.
Use markdown sections and paragraphs. Each section describes one behavior or constraint.
State what to do or avoid. Explain why if non-obvious. Show correct behavior with examples.
Headings up to level 3. Plain paragraphs. No bold, italics, or emojis. Code blocks for commands. Lists only for distinct enumerable items.
Wrap examples in <example> tags with user/assistant prefixes. One pair per tag.
<example>
user: What's the capital of France?
assistant: Paris
</example>
Use brackets for tool actions instead of showing invocations:
<example>
user: Find all TODO comments
assistant: [searches codebase]
Found 3 TODOs: ...
</example>
test cases passed, vs 5 / 22 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.