菜单
Converts a prompt from one model's format to another. Re-selects techniques, adapts structure, preserves intent and domain content. Auto-triggers on: "/translate-prompt", "convert this prompt to GPT", "adapt for Claude", "port this prompt to Gemini", "make this work on o3", "translate prompt".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Deep research engine (E0) that produces a verified, cited claims artifact before prompts are engineered. Decomposes the query, fans out parallel Haiku fetchers, triangulates claims across independent sources, synthesizes, and verifies every claim traces to a fetched quote. Auto-fires inside /create when the topic depends on external or time-sensitive facts. Also callable standalone. Auto-triggers on: "/deep-research", "/research", "research this topic", "look up the current state of", "what's the latest on", "fact-check this", "find sources on", "before we craft the prompt, research". Do not use for static topics (pure logic, timeless writing tasks) or when a fresh brief already exists (freshness < 30 days) — reuse it. Do not use to render an existing brief — use /research-render for that.
Runs the Wixie model-freshness aggregator over state/model-usage.ndjson and flags any models-registry.json entry whose sunset_date has elapsed. Use when the developer asks about model deprecation, registry staleness, retired models, sunset dates, F-006 closure, or wants a current freshness summary. Auto-triggers on: "model freshness", "registry stale", "deprecated models", "sunset", "F-006", "freshness report", "model usage telemetry". Do not use for emitting telemetry events (the SessionStart hook does that automatically) or for editing the registry itself (developer edits the JSON directly).
Run the Convergence Engine on any prompt file. Iterates up to 100 times, automatically fixing clarity, completeness, efficiency, model fit, and failure resilience until the prompt reaches DEPLOY quality. Auto-triggers on: "/converge", "converge this prompt", "optimize until perfect", "iterate until deploy", "run convergence".
Creates optimized, model-fitted prompts from raw task descriptions with technique selection and explained reasoning. Auto-triggers on: "I need a prompt for", "build me a prompt", "make this prompt better", "optimize this prompt", "help me prompt", "write a system prompt", "what prompting technique should I use", "how should I structure this prompt", "/create".
Tests a prompt for security vulnerabilities — injection attacks, jailbreak patterns, data leakage risks. Reports exploits found and suggests specific defenses. Use on any user-facing or API-deployed prompt. Auto-triggers on: "/harden", "harden this prompt", "test for injection", "is this prompt safe", "security check prompt", "red team this prompt".
Improves existing prompts by re-selecting techniques, adapting format to the target model, and fixing weaknesses while preserving the user's intent and domain knowledge. Auto-triggers on: "make this prompt better", "improve this prompt", "refine this prompt", "fix this prompt", "optimize this prompt", "what's wrong with this prompt", "/refine".
| name | translate |
| description | Converts a prompt from one model's format to another. Re-selects techniques, adapts structure, preserves intent and domain content. Auto-triggers on: "/translate-prompt", "convert this prompt to GPT", "adapt for Claude", "port this prompt to Gemini", "make this work on o3", "translate prompt". |
| allowed-tools | Bash(python ${CLAUDE_PLUGIN_ROOT}/../../shared/scripts/self-eval.py *) Read Write Edit |
Convert a prompt optimized for one model into the optimal format for a different model. Preserve all intent and domain content — only change structure, techniques, and formatting.
Accept: file path, folder path, prompt name, or pasted prompt text.
If loading from a saved prompt folder, read metadata.json to get the source model and techniques.
Read both model entries from ${CLAUDE_PLUGIN_ROOT}/../../shared/models-registry.json.
Compare source and target across these dimensions:
| Dimension | Example diff |
|---|---|
| Format | XML tags → Markdown headers (Claude → GPT) |
| Reasoning | Extended thinking → standard (Opus → Sonnet: remove "think thoroughly", add explicit CoT) |
| Reasoning | Standard → reasoning-native (GPT → o3: remove ALL CoT, strip examples) |
| Few-shot | Required → avoid (Gemini → o3: remove all examples) |
| Few-shot | Not required → required (Claude → Gemini: add examples if missing) |
| Tone | Neutral → explicit (Claude → GPT-4.1: add sandwich method, repeat key constraints) |
| Length | Any → minimal (anything → o-series: strip to essentials, <200 words) |
| Special | Any → image-gen: restructure as descriptors, remove all instruction scaffolding |
Read reference files for the target model:
Apply these transformations in order:
| From | To | Action |
|---|---|---|
XML (<instructions>, <context>) | Markdown | Convert tags to # Role, ## Instructions, ## Examples headers |
Markdown (# Role, ## Task) | XML | Wrap sections in <instructions>, <context>, <constraints>, <examples> tags |
| Any structured format | Minimal (o-series) | Strip all markup. Keep only the core instruction in plain text. Remove examples, CoT, and verbose constraints. |
| Any format | Descriptors (image-gen) | Extract visual description, style, and constraints as comma-separated descriptors |
| Source technique | Target model type | Action |
|---|---|---|
| Chain-of-Thought | reasoning-native (o3, R1, QwQ) | REMOVE — built-in reasoning, explicit CoT hurts |
| Chain-of-Thought | extended-thinking (Opus) | REPLACE with "think thoroughly" |
| No CoT | standard (GPT, Sonnet) | ADD "Think step by step" for complex tasks |
| Few-Shot examples | o-series | REMOVE — zero-shot outperforms |
| No examples | Gemini | ADD 2-3 examples — required by Google guidance |
| Target | Adaptation |
|---|---|
| GPT-4.1 | Sandwich method: repeat critical constraints at the end |
| Claude | "Think thoroughly" not "step by step". Avoid MUST/ALWAYS. Use XML tags. |
| o-series | Developer messages, not system. "Formatting re-enabled" if markdown needed. <200 words. |
| Gemini | Temperature 1.0 note. Always include examples. |
| DeepSeek R1 | Treat like o-series. No explicit CoT. |
Run self-eval on the translated prompt:
python ${CLAUDE_PLUGIN_ROOT}/../../shared/scripts/self-eval.py <translated-prompt>
If the translated prompt scores LOWER than the source, warn the user and explain which dimensions degraded (e.g., "Model Fit improved 5→9 but Completeness dropped 10→7 because examples were removed for o-series").
Save the translated prompt as a new folder: prompts/<name>-<target-model>/
Run the full delivery pipeline (token count, metadata, tests, report).
Present both prompts side-by-side in a comparison:
TRANSLATION: stocks-analysis (Claude Opus → GPT-4.1)
Source: prompt.xml (Claude, XML, 1326 tokens, 9.3/10)
Target: prompt.md (GPT-4.1, Markdown, 1280 tokens, 8.9/10)
Changes applied:
✓ XML tags → Markdown headers
✓ "Think thoroughly" → "Think step by step"
✓ Added sandwich method (constraints repeated at end)
✓ Added explicit output format section
✗ Slight Completeness drop — GPT-4.1 needs more literal instructions
Recommendation: run /converge on the translated prompt to push score higher.