Complete methodology for preventing AI hallucinations. Use when accuracy is critical and you need Claude to verify before claiming, cite before asserting, and admit uncertainty instead of guessing.
中文简介正在准备中。当前显示原文说明。
Trace bugs backward to their origin, not just symptoms
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
TRACE BACKWARD TO THE BUG'S ORIGIN
Don't patch symptoms - fix the disease
SYMPTOM
↓ Where does this value come from?
IMMEDIATE CAUSE
↓ What set that to wrong value?
DEEPER CAUSE
↓ Why did that happen?
ROOT CAUSE ← Fix HERE
Q: What exactly went wrong?
→ Document the exact symptom
→ Note file, line, error message
Q: Where does this value/state come from?
→ Find the direct source
→ Check: is THIS the bug or just a carrier?
Q: What set THAT to the wrong value?
→ Follow the data flow backward
→ Each step: is this the origin or just passing it along?
SIGNS YOU'VE FOUND ROOT:
- Logic error in requirements interpretation
- Initial state incorrectly set
- Wrong assumption in algorithm
- Missing validation at entry point
- Broken invariant at creation
You've found root cause when:
You're still at symptoms if:
| Bug Pattern | Root Usually At |
|---|---|
| Wrong value displayed | Data calculation, not display code |
| Null pointer | Where object should've been created |
| Off-by-one | Loop bounds definition, not body |
| Race condition | Shared state design, not specific access |
| Invalid state | State transition logic, not current handler |
❌ Symptom Patching
// Bad: Fix where error happens
if (value === undefined) value = defaultValue
// Good: Find why value is undefined
❌ Guard Stacking
// Bad: Add guards everywhere
if (x) if (y) if (z) doThing()
// Good: Ensure x, y, z correct at source
❌ Quick Fix Pressure
"Just make it work" → Technical debt
Better: "5 more minutes tracing saves 5 hours debugging later"
After identifying root cause:
Use with:
assumption-checker → Was root cause an assumption?dependency-validator → Was root cause a broken dependency?pre-action-verifier → Verify root cause fix before applying还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Complete methodology for preventing AI hallucinations. Use when accuracy is critical and you need Claude to verify before claiming, cite before asserting, and admit uncertainty instead of guessing.
中文简介正在准备中。当前显示原文说明。
Transform rough ideas into solid designs through structured questioning
中文简介正在准备中。当前显示原文说明。
Canvas API Connection Skill
中文简介正在准备中。当前显示原文说明。
Forces every code-related claim to include file:line citations. No citation means the claim must be verified or removed. Use when building documentation, writing code reviews, or any output that others will rely on.
中文简介正在准备中。当前显示原文说明。
Assign confidence scores (0-100) to every claim in a response. Helps users understand which parts are verified facts and which are educated guesses. Use when the user needs to know how much to trust each part of the answer.
中文简介正在准备中。当前显示原文说明。
Forces Claude to extract exact quotes from files before making claims about them. Prevents hallucinations by grounding every statement in actual text from the codebase. Use when working with large files or unfamiliar code.
中文简介正在准备中。当前显示原文说明。