7 Claude prompt prefixes that change reasoning

admin By admin
April 14, 2026
0

Prompt

ULTRATHINK — Maximum reasoning depth. Claude thinks longer, catches edge cases it normally misses. L99 — Kills hedging. /ghost — Strips AI writing patterns. /skeptic — Challenges your premise before answering. PERSONA (with specificity) — "Senior M&A attorney at a top-100 firm, 20 years, skeptical of boilerplate" produces fundamentally different output than just asking a legal question. /debug — Forces Claude to find the bug instead of rewriting your code. OODA — Structures response as Observe-Orient-Decide-Act.

Description

ULTRATHINK — Maximum reasoning depth. Claude thinks longer, catches edge cases it normally misses. L99 — Kills hedging. /ghost — Strips AI writing patterns. /skeptic — Challenges your premise before answering. PERSONA (with specificity) — "Senior M&A attorney at a top-100 firm, 20 years, skeptical of boilerplate" produces fundamentally different output than just asking a legal question. /debug — Forces Claude to find the bug instead of rewriting your code. OODA — Structures response as Observe-Orient-Decide-Act.

Example Outputs

The 7 prompt prefixes that actually change Claude's reasoning and output are: ULTRATHINK - Prompts Claude to think more deeply, consider more edge cases, and provide more nuanced recommendations. This was particularly useful for architecture questions. L99 - Removes hedging language, leading to more decisive, action-oriented responses. Good for making actual decisions. /ghost - Strips AI writing patterns like em-dashes and 'it's worth noting', reducing detection of AI-generated text. /skeptic - Challenges the premise of the question, rather than just optimizing the proposed solution. This prevented building the wrong thing. PERSONA (with specificity) - Using a detailed persona with stated biases and experience changes the fundamental nature of the response, versus a generic prompt. /debug - Forces Claude to identify and explain bugs in code, rather than just rewriting it. OODA - Structures the response using the military decision framework of Observe, Orient, Decide, Act. Useful for time-pressured decisions.
Category: Chatbots
Tested with: Claude (Anthropic)

Comments

Please log in to comment.

No comments yet. Be the first to comment!

Share Feedback