optimize-prompt

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Optimize Prompt

优化Prompt

Prompt optimization pipeline. Given prompt P, improve via critique then compression, with semantic equivalence checks after each stage.

Prompt优化流水线：给定Prompt P，先通过评估再进行压缩优化，每个阶段结束后都会进行语义等价性校验。

Pipeline

流水线

P → [Input Analysis] → [Critique] → P' → [Equiv Check 1] → [Compress] → P'' → [Equiv Check 2] → Output P''

Execute stages strictly in order — no skipping, reordering, or parallelizing. Each stage passes (pipeline continues) or fails (pipeline terminates with explanation).

P → [Input Analysis] → [Critique] → P' → [Equiv Check 1] → [Compress] → P'' → [Equiv Check 2] → Output P''

必须严格按顺序执行各个阶段，不得跳过、调整顺序或并行执行。每个阶段要么通过（流水线继续执行），要么失败（流水线终止并给出解释）。

Stage 0: Input Analysis

阶段0：输入分析

Examine P:

Does P take input? Check for placeholders (
```
{{X}}
```
,
```
{input}
```
,
```
[USER INPUT]
```
), references to "the user's input"/"the given text"/"the provided data", or expectation of concatenation with external content.
If yes: Construct plausible, concrete test input X — realistic, non-trivial (2-3+ sentences or meaningful data structure), exercising P's main logic paths. If P has branching conditions, X triggers the primary branch. If P produces structured output, X requires all output fields. If P has validation conditions, X passes validation. State X explicitly.
If no: P is self-contained. Set X = ∅. Equivalence checks compare outputs of P directly.

Display:

**Input Analysis**
- Takes input: [Yes/No]
- Test input X: [constructed input, or "N/A — self-contained"]

检查P的属性：

P是否需要接收输入？ 检查是否存在占位符（
```
{{X}}
```
、
```
{input}
```
、
```
[USER INPUT]
```
）、是否提及「用户输入」/「给定文本」/「提供的数据」，或者是否需要和外部内容拼接使用。
如果需要输入：构造合理、具体的测试输入X，要求真实、非琐碎（至少2-3句话或者有意义的数据结构），能够覆盖P的主要逻辑路径。如果P有分支条件，X要触发主分支；如果P生成结构化输出，X要覆盖所有输出字段；如果P有校验条件，X要满足校验规则。需要明确写出X的内容。
如果不需要输入：说明P是自包含的，设置X = ∅，等价性校验直接对比P的输出即可。

输出内容：

**Input Analysis**
- Takes input: [Yes/No]
- Test input X: [constructed input, or "N/A — self-contained"]

Stage 1: Critique

阶段1：评估

Apply think-critically methodology to P:

Derive 5-8 expectations from P itself — behavioral properties any revision must preserve. Frame as testable statements (e.g., "Produces JSON output", "Rejects off-topic queries").
Evaluate P against expectations. Per expectation: confidence (0-100%) with concise rationale referencing specific text in P.
Propose fixes for expectations with confidence < 95%. Each fix: exact text to add, remove, or replace.
Produce P' — revised prompt with all fixes applied. Complete text, not a diff.

Display:

**Stage 1: Critique**

| Expectation | Confidence | Rationale |
|---|---|---|
| ... | ...% | ... |

**Overall Score: [average]%**

**Fixes Applied:**
1. [Fix with exact text changes]
...

---
> **P' (Revised Prompt):**
---
[Full text of P']
---

If all expectations >= 95%, set P' = P and note "No fixes needed."

对P运用批判性思维方法进行优化：

从P本身推导5-8条预期规则——所有修改版本都必须保留的行为属性，表述为可测试的语句（例如「生成JSON格式输出」、「拒绝无关主题的查询」）。
对照预期规则评估P：针对每条预期给出置信度（0-100%），并给出引用P中具体内容的简短理由。
为置信度低于95%的预期提出修复方案：每个方案要明确写出需要添加、删除或替换的具体文本。
生成P'——应用所有修复后的完整Prompt内容，不要只给出差异对比。

输出内容：

**Stage 1: Critique**

| Expectation | Confidence | Rationale |
|---|---|---|
| ... | ...% | ... |

**Overall Score: [average]%**

**Fixes Applied:**
1. [Fix with exact text changes]
...

---
> **P' (Revised Prompt):**
---
[Full text of P']
---

如果所有预期的置信度都≥95%，则设置P' = P，并标注「无需修复」。

Stage 2: Equivalence Check 1

阶段2：等价性校验1

Verify P and P' produce essentially the same output on X.

Procedure:

Simulate running P on X. Describe expected output (2-4 sentences): structure, content, tone, key features.
Simulate running P' on X. Describe expected output (2-4 sentences).
Equivalent if all hold:
- (a) Output structure identical (sections, format, ordering)
- (b) Factual/decisional content identical — nothing added, removed, or altered
- (c) Differences limited to: wording improvements, added specificity, stronger constraint enforcement
- (d) User expecting P's behavior would accept output without noticing intent change
- If (a)-(c) hold but (d) uncertain, default YES.

If X = ∅, compare standalone outputs.

Display:

**Stage 2: Equivalence Check 1**

- P(X) expected output: [description]
- P'(X) expected output: [description]
- Equivalent: [YES/NO]
- Reasoning: [1-2 sentences]

If NO: Terminate:

**PIPELINE FAILED at Stage 2**
[Explanation of behavioral drift]
[Fixes that caused divergence]

If YES: Proceed to Stage 3.

验证P和P'在输入X下生成的输出本质一致。

执行流程：

模拟在X上运行P，描述预期输出（2-4句话）：包括结构、内容、语气、核心特征。
模拟在X上运行P'，描述预期输出（2-4句话）。
满足以下所有条件则判定为等价：
- (a) 输出结构完全一致（章节、格式、顺序）
- (b) 事实/决策内容完全一致，没有新增、删除或修改
- (c) 差异仅局限于：措辞优化、新增明确性说明、更强的约束执行
- (d) 预期P行为的用户可以接受该输出，不会感知到意图变化
- 如果满足(a)-(c)但(d)不确定，默认判定为等价。

如果X = ∅，直接对比独立运行的输出。

输出内容：

**Stage 2: Equivalence Check 1**

- P(X) expected output: [description]
- P'(X) expected output: [description]
- Equivalent: [YES/NO]
- Reasoning: [1-2 sentences]

如果判定为不等价：终止流水线：

**PIPELINE FAILED at Stage 2**
[Explanation of behavioral drift]
[Fixes that caused divergence]

如果判定为等价：进入阶段3。

Stage 3: Compress

阶段3：压缩

Apply compress-prompt methodology to P' (lossless mode):

Target 10-30% token reduction, 100% semantic retention.
Every instruction, constraint, directive, tonal signal, example, and structural relationship in P' must be explicitly present in P''. Nothing left to inference.
Allowed: remove filler, collapse redundancy, tighten syntax, merge duplicates, normalize structure.
Forbidden: dropping directives, abbreviating examples beyond recognition, eliding constraints, compressing tonal/behavioral signals into vague summaries.
Produce P'' and a directive map (each P' directive → P'' counterpart).

Display:

**Stage 3: Compress**

---
> **P'' (Compressed Prompt):**
---
[Full text of P'']
---

**Directive Map:**
| # | Original directive (P') | Compressed counterpart (P'') |
|---|---|---|
| 1 | [directive from P'] | [text in P''] |
| ... | ... | ... |

**Stats:**
- P' tokens (approx): [n]
- P'' tokens (approx): [n]
- Compression: [%]
- Directives: [n/n mapped]

对P'应用Prompt压缩方法（无损模式）：

目标是减少10-30%的Token量，100%保留语义。
P'中的每一条指令、约束、规则、语气提示、示例、结构关系都必须明确出现在P''中，不得依赖推理补全。
允许的操作：删除填充内容、合并冗余信息、精简语法、合并重复内容、规范结构。
禁止的操作：删除指令、过度缩写示例导致无法识别、省略约束、将语气/行为信号压缩为模糊的总结。
生成P''和指令映射表（每条P'的指令对应P''中的对应内容）。

输出内容：

**Stage 3: Compress**

---
> **P'' (Compressed Prompt):**
---
[Full text of P'']
---

**Directive Map:**
| # | Original directive (P') | Compressed counterpart (P'') |
|---|---|---|
| 1 | [directive from P'] | [text in P''] |
| ... | ... | ... |

**Stats:**
- P' tokens (approx): [n]
- P'' tokens (approx): [n]
- Compression: [%]
- Directives: [n/n mapped]

Stage 4: Equivalence Check 2

阶段4：等价性校验2

Verify P' and P'' produce essentially the same output on X.

Procedure: Same as Stage 2, comparing P' and P''.

Display:

**Stage 4: Equivalence Check 2**

- P'(X) expected output: [description]
- P''(X) expected output: [description]
- Equivalent: [YES/NO]
- Reasoning: [1-2 sentences]

If NO: Terminate:

**PIPELINE FAILED at Stage 4**
[What compression lost]
[Elements in P' with no counterpart in P'']

If YES: Proceed to output.

验证P'和P''在输入X下生成的输出本质一致。

执行流程： 和阶段2一致，对比P'和P''的输出。

输出内容：

**Stage 4: Equivalence Check 2**

- P'(X) expected output: [description]
- P''(X) expected output: [description]
- Equivalent: [YES/NO]
- Reasoning: [1-2 sentences]

如果判定为不等价：终止流水线：

**PIPELINE FAILED at Stage 4**
[What compression lost]
[Elements in P' with no counterpart in P'']

如果判定为等价：进入最终输出环节。

Final Output

最终输出

**PIPELINE SUCCEEDED**

Your optimized prompt (P''):

---
> **BEGIN OPTIMIZED PROMPT**
---
[Full text of P'']
---
> **END OPTIMIZED PROMPT**
---

**Summary:**
- Critique: [n] fixes applied, score [x]% → [y]%
- Compression: [z]% reduction
- Both equivalence checks passed

**PIPELINE SUCCEEDED**

Your optimized prompt (P''):

---
> **BEGIN OPTIMIZED PROMPT**
---
[Full text of P'']
---
> **END OPTIMIZED PROMPT**
---

**Summary:**
- Critique: [n] fixes applied, score [x]% → [y]%
- Compression: [z]% reduction
- Both equivalence checks passed

Edge Cases

边界情况

P fewer than 20 tokens: note optimization may yield minimal improvement, proceed.
P already optimal (all expectations >= 95%, compression < 10%): state "Prompt is already well-optimized", return P unchanged.
Adversarial or self-referential P: evaluate literally, note observation.
Compression < 10% reduction without dropping directives: skip compression, set P'' = P', note "Compression skipped — prompt already dense."
P' more than 50% longer than P: flag "Significant expansion", verify in Stage 2 that expansion only adds guardrails/specificity.
P contains code blocks: preserve verbatim during critique and compression unless fix explicitly targets code content.

FIRST-TOKEN CONSTRAINT: Response must begin with "## Optimize Prompt". No greetings, preambles, commentary, or blank lines before it. Overrides default conversational behavior.

P的Token数少于20：标注优化可能带来的提升很小，继续执行流程即可。
P已经是最优状态（所有预期置信度≥95%，可压缩比例<10%）：标注「Prompt已经完成良好优化」，原样返回P。
对抗性或自引用的P：按字面内容评估，标注相关观察。
不删除指令的前提下压缩比例<10%：跳过压缩步骤，设置P'' = P'，标注「跳过压缩——Prompt已经非常精简」。
P'的长度比P长50%以上：标注「大幅扩容」，在阶段2中验证扩容仅新增了防护规则/明确性说明。
P包含代码块：在评估和压缩阶段原样保留，除非修复方案明确针对代码内容。

首Token约束：响应必须以「## Optimize Prompt」开头，前面不得有问候语、前言、注释或空行，覆盖默认的对话行为。

Input

输入

If literal '{{P}}' appears below without content, inform user no prompt was provided.

Prompt to optimize:

如果下方出现无内容的字面量 '{{P}}'，请告知用户未提供待优化的Prompt。

待优化Prompt：