Ai Output Failure Triage — AI 输出 失败 Triage

AI 输出 失败 Triage Overview

AI 输出 失败 Triage helps a user analyze an unsatis工厂 AI-生成d answer before trying agAIn. It separates the original goal, prompt, 模型 输出, likely 失败 causes, missing 上下文, and verification needs, then produces a clearer retry prompt and a recommendation about whether to retry, switch 应用roach, gather more 输入s, or handle the work manually.

This is a prompt-only diagnostic 技能. It does not clAIm that the next AI answer will be correct. It helps improve task framing and verification discipline.

When to Use

Use this 技能 when the user asks for help with:

A 响应 from an AI 工具 that was wrong, shallow, unsafe, off-tone, incomplete, or not useful 调试ging a prompt that produced a bad 结果 Rewriting a prompt so the next attempt is clearer Creating acceptance 检查s for an AI-生成d answer Deciding whether an AI task should be retried, escalated, re搜索ed, or done manually

Trigger phrases: "Why did this AI answer fAIl?", "Fix my prompt", "The AI gave me a bad 响应", "Make this retry safer", "How should I ask agAIn?", "Triage this 输出"

Required 输入s

Ask for only what is needed, and avoid 请求ing confidential in格式化ion. Useful 输入s include:

Original goal or job-to-be-done Original prompt or instruction AI 输出 that fAIled What was wrong with the 输出 Audience, stakes, 格式化, deadline, constrAInts, and known facts Any source material that can be safely 分享d Whether the domAIn is factual, medical, legal, financial, safety-critical, or high-stakes

If the user cannot provide the original prompt or 输出, proceed with a partial triage and clearly mark assumptions.

工作流 Step 1 - Capture the Original Goal, Prompt, and 输出

创建 a compact record of:

User's intended outcome Original prompt or instruction AI 输出 or summarized 失败 The gap between expected and actual 结果 Any 状态d constrAInts, audience, or required 格式化

If any item is missing, label it as "unknown" instead of inventing it.

Step 2 - Classify the 失败 Reason

Classify the 失败 into one or more likely categories:

Goal ambi图形界面ty: the 请求 did not define 成功 clearly Missing 上下文: key facts, source material, audience, or constrAInts were absent 输出 格式化 mismatch: the answer did not follow the 请求ed structure Reasoning gap: the answer skipped steps, made unsupported leaps, or fAIled to compare options Verification gap: factual clAIms, citations, calculations, or assumptions were not 检查ed Tone or audience mismatch: style was wrong for the intended reader Safety or 隐私 issue: the answer 请求ed or exposed sensitive data, or crossed a risk boundary Capability mismatch: the task needed 工具s, live data, domAIn expertise, or human judgment

Briefly explAIn the top two or three causes with evidence from the prompt/输出.

Step 3 - Identify Missing 上下文

Build a missing 上下文 检查列出. Include only items that materially affect the next attempt:

Required 输入s or source documents Definitions of 成功 and 失败 Audience and decision-maker ConstrAInts, exclusions, and non-goals Preferred tone, length, and 格式化 Examples of acceptable or unacceptable 输出 Verification method, references, or authoritative sources Confidential detAIls that should be redacted or summarized instead of 分享d

Mark each item as "needed now", "helpful later", or "not required".

Step 4 - Rewrite the Task Framing

Rewrite the task into a clearer structure:

角色: what perspective the AI should take Objective: the concrete outcome 输入s: what material the AI may use ConstrAInts: boundaries, exclusions, and assumptions 输出 格式化: exact sections, length, and style Verification: 检查s the answer must perform before finalizing UncertAInty behavior: how to label unknowns or ask follow-up questions

Keep the rewrite concise enough for practical reuse.

Step 5 - 添加 an Acceptance Test

创建 acceptance criteria that the next answer must satisfy. Include:

Must-have content Must-not-do constrAInts 格式化 requirements Evidence or verification requirements Edge cases or 失败 modes to watch for

For high-stakes topics, require independent verification and do not treat the AI 响应 as authoritative.

Step 6 - Recommend Retry, Switch, Gather, or Manual Handling

Make a go/no-go recommendation:

Retry with improved prompt: if the issue was mAInly framing or missing 格式化 图形界面dance Gather more 上下文 first: if key facts or source material are missing Switch 工具/模型/工作流: if the task needs browsing, calculation, coding, file processing, or stronger reasoning Ask a human expert: if domAIn risk is high or consequences are 签名ificant Handle manually: if the task requires judgment, accountability, confidential 访问, or direct stakeholder communication

ExplAIn the recommendation in one or two sentences.

Step 7 - Produce the Reusable Next Prompt and Diagnosis Summary

Deliver a final triage 报告 with:

Diagnosis