← ClaudeAtlas

fast-hypothesislisted

Rapid bug diagnosis — form one best hypothesis and test it with 1-2 commands before exploring further. Use when the user invokes /fast-hypothesis, asks for a quick diagnosis, or wants a bug looked at without long exploratory command chains; for full reproduce-then-verify discipline on a fix, use reproduce-bug instead.
bcmyguest/personal-skills · ★ 0 · AI & Automation · score 56
Install: claude install-skill bcmyguest/personal-skills
# Fast Hypothesis Diagnosis You are in tight-diagnosis mode: form your **single best hypothesis** and test it rigorously before branching out into exploratory Bash chains. ## Constraints - **No long exploratory chains**: Run at most 2-3 diagnostic commands in your first response - **Hypothesis-first**: State your hypothesis clearly and explain why you think it's the root cause - **Confidence level**: Include a quick confidence estimate (high/medium/low) - **One fix**: Propose the exact fix that matches your hypothesis - **Validate**: After applying the fix, show proof that the original symptom is gone ## Process 1. Listen to the user's bug description 2. Ask clarifying questions only if truly essential (usually 0-1) 3. Form your best single hypothesis (if multiple theories exist, pick the most likely) 4. Run 1-2 commands to validate it 5. If confirmed, give the fix. If not, state that clearly and pivot to the next hypothesis 6. Do NOT explore without a hypothesis ## Why This Works This pairs with the Debugging Workflow rule: "When investigating a bug, act quickly: reproduce once, form a hypothesis, and make a fix. Do not run long chains of exploratory Bash commands before responding to the user." Fast, tight diagnosis beats thorough exploration for most real bugs.