home/productivity/sampling-base

Sampling Base

GPTClaudeDeepSeek··1,283 copies·updated 2026-07-14
sampling-base.prompt
# PRISM Sampling Prompt Guidelines

When designing prompts for PRISM N-Sampling, the goal is to maximize the useful variance between instances.

## Principles

**Be specific about the task, not the approach**

❌ "Use a structured approach to analyze..."
✅ "Analyze X and give me actionable conclusions"

Specifying the approach constraints the trajectory space. Let the model find its own path.

**Include relevant context, avoid implicit constraints**

Include the facts the model needs. Don't add stylistic constraints ("be concise", "use bullet points") — those reduce variance. Format instructions belong in the meta-agent prompt, not here.

**Open-ended > Closed-ended**

Questions with one correct answer don't benefit from N-sampling. The gain is maximal when the problem has multiple valid approaches or when the solution space is large.

## Template Structure

when to use it

Community prompt sourced from the open-source GitHub repo Mnemoclaw/prism-framework (MIT). A "Sampling Base" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

productivitycommunitydeveloper

source

Mnemoclaw/prism-framework · MIT