Attack the premise

SkillAI & models

Helps your agent question shared assumptions behind failed fixes instead of repeating them.

Available today. Use it from your connected AI after setup.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the Attack the premise skill

About this skill

Apply when repeated fixes sharing an assumption fail. State the assumption and choose an observation that can challenge it before trying another fix that depends on it.

What this skill tells your AI

The instructions your AI receives, as published by michael-denyer/pstack-claude in plugins/pstack/skills/principle-attack-the-premise/SKILL.md and read by ahel’s review.

When two or more fixes sharing an assumption fail the same gate, write down what they assumed. Before another fix depends on it, choose an observation or experiment that can challenge it.

Match the experiment to the hypothesis. Two timeout increases may assume requests reach the intended destination. Check the actual destination before increasing the timeout again.

When the hypothesis concerns uneven allocation, ownership, or assignment, count resources or work per actor. Measure the distribution and trace the assignment per Fix Root Causes. Recommend redistribution, rotation, or reassignment only when evidence connects that assignment to the failure. An even distribution weakens the imbalance hypothesis; it does not rule out a shared defect. Every worker can leak at the same rate.

Keep a rerunnable experiment when useful, per Build the Lever. Do not turn this into a mandatory census for unrelated bugs.

This principle questions an assumption about the existing system. Redesign from First Principles rebuilds a design around a new requirement.

Signals

GitHub stars
667
Forks
78
Last commit
Sep 2026
Advanced
Item type
skill
Key
principle-attack-the-premise-michael-denyer
Source
github.com/michael-denyer/pstack-claude