Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add IchenDEV/prompt-optimizer-plugins --skill optimize-deepseek-v4-promptsgit clone --depth 1 https://github.com/IchenDEV/prompt-optimizer-pluginsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/ichendev/prompt-optimizer-plugins/optimize-deepseek-v4-prompts)<a href="https://agentmods.dev/skills/ichendev/prompt-optimizer-plugins/optimize-deepseek-v4-prompts"><img src="https://agentmods.dev/badge/skills/ichendev/prompt-optimizer-plugins/optimize-deepseek-v4-prompts/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ichendev/prompt-optimizer-plugins/optimize-deepseek-v4-prompts"><img src="https://agentmods.dev/badge/skills/ichendev/prompt-optimizer-plugins/optimize-deepseek-v4-prompts.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00151 | $0.01494 |
| Opus 5 | $0.00076 | $0.00747 |
| Sonnet 5 | $0.00030 | $0.00299 |
| Haiku 4.5 | $0.00015 | $0.00149 |
Grade A, and why
optimize-deepseek-v4-prompts scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Optimize DeepSeek-V4 Prompts
Turn rough ideas and existing prompt stacks into copy-ready prompts for DeepSeek-V4 Pro or Flash. Preserve the user's intent and constraints, ask for genuinely blocking information, and distinguish paper-reported behavior from prompt-design inference.
Use the technical-report basis
Read references/official-guidance.md before applying model-specific advice. The source is a DeepSeek-AI technical report, not a general prompt-engineering guide, so keep claims bounded to what it reports and the explicitly labeled optimizer implications. Fetch the live report or current platform documentation before making claims about released model IDs, API fields, availability, limits, or pricing.
Follow the workflow
1. Capture the prompt contract
Identify:
- the intended outcome, audience, language, and risk level;
- whether the target is DeepSeek-V4 Pro, Flash, or an unspecified V4 variant;
- the inputs, reference corpus, search access, and real tools the model will receive;
- facts and constraints to preserve, especially complex writing constraints;
- required output content, format, evidence, verification, and completion criteria;
- whether runtime reasoning mode or an end-to-end API request was actually requested.
Treat these as diagnostic dimensions, not mandatory headings. Preserve system, developer, user, and tool-description boundaries when the user supplies a layered prompt stack.
2. Apply the clarity gate
Ask a question only when two reasonable answers would materially change the outcome, source boundary, model variant, tool permissions, output schema, or success criteria.
Treat these gaps as blocking by default:
- a vague topic and generic verb with no identifiable deliverable or audience;
- an unspecified corpus, search boundary, or tool set whose choice changes the answer;
- an external action without clear authorization;
- conflicting complex constraints or an undefined strict schema;
- a request for variant-specific or runtime advice when Pro versus Flash or latency versus quality materially changes the recommendation.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 91 lines · 151 tokens per session scan A 7da2b714b6f2
optimize-deepseek-v4-prompts is a skill published in the GitHub repository IchenDEV/prompt-optimizer-plugins (6 stars, last pushed 3d ago), licensed MIT. It adds 151 tokens to every session and 1,494 once invoked, about $0.0008 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
instructor
Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor - battle-tested structured output library.
build-dotnet-agent-service
Build a local-first F# or C# Semantic Kernel agent service with explicit tools, model capability checks, evaluation fixtures, and draft-before-write promotion.
instructor
Extract structured data from LLM responses with Pydantic validation, retry failed extractions automatically, parse complex JSON with type safety, and stream partial results with Instructor - battle-tested structured output library.
analyze-prompt
Use when the user wants to review, score, or improve how they write prompts; when coaching prompt quality; or when tracking prompting mistakes over a session. Keywords: prompt analysis, prompt coaching, how should I ask, rate my prompt, am I prompting well, prompt statistics, prompt score.
prompt-mini
Forges weak Claude Code prompts into structured, credit-saving, framework-aware prompts. Use when prompts are vague, missing scope, or mention a stack without clear task definition.
prompt-report
Use when the user wants an end-of-session summary of their prompting: overall stats, best and worst prompts, biggest recurring mistake, improvement curve, and personalized recommendations. Keywords: prompt report, session summary, how did I do, my prompting stats, prompt scorecard, prompting progress.