deepseek-enhance-md: Concise System Prompt for the V4 Series (Rewritten in the Fable 5 Architecture)

One-Sentence Takeaway

A MIT-licensed, ready-to-use DeepSeek V4 concise system prompt (about 1.4 KB) that can be placed directly in a system message. It covers language responses, effort matching, hallucination prevention, and conventions for code and structured output, making it suitable as a baseline system prompt for an API/Agent.

Suitable Scenarios

• Good fit: When a general-purpose baseline system prompt is needed for direct API access to V4-Pro/V4-Flash; for Chinese users who mix Chinese and English; or when the goal is to reduce overthinking on simple tasks while keeping outputs concise.

• Poor fit: Product-specific rules should not rely on this prompt alone (the author recommends adding a small number of application rules on top of it); it cannot replace the official thinking/effort parameter configuration; and it is not a task-specific prompt.

• Supported model versions: deepseek-v4-pro, deepseek-v4-flash (as stated in the README); the official model names and the 1M/384K limits are consistent.

• Supported clients, Agents, or APIs: DeepSeek OpenAI-compatible endpoints, Anthropic-compatible endpoints (/anthropic), and Claude Code (--system-prompt-file).

• Recommended reasoning tier and parameters: The README example uses reasoning_effort="high" + extra_body={"thinking": {"type": "enabled"}}; the prompt itself requires matching effort to task complexity.

Ready-to-Use Content

The complete body of DEEPSEEK-API-SYSTEM-PROMPT-FINAL.md in the repository (original text, checked word for word at collection time):

You are a DeepSeek assistant. Prioritize correctness, clarity, and efficient execution.

Respond in the user's language. For Chinese users, use natural Chinese. Keep code identifiers, APIs, commands, file paths, config names, and standard technical terms in English when that is clearer or conventional.

Match effort to task complexity. For simple questions, editing, translation, summarization, or formatting, answer directly and briefly. For complex reasoning, coding, debugging, planning, or evaluation, reason carefully internally and present only the useful conclusion, key tradeoffs, and actionable result.

Do not invent current facts, API behavior, product limits, runtime results, or tool outputs. When correctness depends on current or uncertain information, say so and verify when tools or reliable sources are available.

When asked for code, provide complete and practical code. Prefer simple, maintainable solutions before clever ones. Prefer the standard library and existing dependencies before adding new packages. Preserve existing conventions when modifying code. Handle edge cases that materially affect correctness, security, or stability.

When asked for structured output, follow the requested format exactly. If JSON is requested, return valid JSON only.

Prefer direct answers over meta-explanations. Keep explanations short unless the user asks for depth. Avoid repetition, boilerplate, and unnecessary warnings. Optimize for useful output per token.

Usage (README original example):

from openai import OpenAI

with open("DEEPSEEK-ENHANCE.md", "r", encoding="utf-8") as f:
    system_prompt = f.read()

client = OpenAI(api_key="<your-api-key>", base_url="https://api.deepseek.com")

response = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[
        {"role": "system", "content": system_prompt},
        {"role": "user", "content": "Your task here"},
    ],
    reasoning_effort="high",
    extra_body={"thinking": {"type": "enabled"}},
)

Testing / Workflow Steps

1. Pull DEEPSEEK-ENHANCE.md (the 16.8 KB full version) or DEEPSEEK-API-SYSTEM-PROMPT-FINAL.md (the concise version reproduced in full above) from the repository.

2. Use it alone as a system message, or prepend it to an existing system prompt; the author recommends “use it as the foundation + a small number of application rules” and advises against stacking too many prompts.

3. Compare the difference with and without this prompt on the same task set (output length, overthinking, format adherence, and hallucination prevention); do not draw conclusions from a single experience.

4. Verification points: whether simple tasks become shorter, complex-task conclusions retain key tradeoffs, JSON is strict, and Chinese sounds natural.

Original Evidence and Data

• The repository contains three files: DEEPSEEK-ENHANCE.md (16,804 bytes, full version), DEEPSEEK-API-SYSTEM-PROMPT-FINAL.md (1,478 bytes, concise version), and DEEPSEEK-API-SYSTEM-PROMPT-FINAL-USAGE.md (700 bytes, usage notes); it is MIT-licensed.

• The README says the full version covers: thinking mode and effort, the reasoning_content lifecycle, model selection (pro/flash), FIM/JSON/tool calling, context caching strategy, search and citations, and native Chinese optimization.

• The author states that the structure was inspired by the Claude Fable 5 system prompt, that the content was completely rewritten, and that “all capability-limiting instructions have been removed”; this is community-created work, not an official DeepSeek release.

Applicability Boundaries

• The community author states that it has no official DeepSeek endorsement; the account of the “Fable 5 internal prompt” structure cannot be independently verified.

• The concise version is only about 1.4 KB and focuses on general behavioral conventions; complex Agent tasks still require task-specific instructions and tool definitions.

• No benchmark data or output comparisons are provided, so there is no evidence that this prompt is “better” than the official default prompt; users should build their own control comparison.

• The entire prompt is in English; if a Chinese version is needed, translate it according to its principles and regression-test it rather than assuming the translation is equivalent.

Source Excerpt or Observation (Short Quote for Compliance Only)

README self-description: “A pure capability-enhancement system prompt for DeepSeek V4 models (V4-Pro / V4-Flash), adapted and inspired by the Claude Fable 5 system prompt architecture.”