古诗词生图器

SKILL内容创作
内容创作 未标注权限 许可证不明确
安装这个技能

面向 Codex 桌面端 的独立诗词生图技能:搜索或随机选择中国古诗词,提炼意象、匹配原创视觉风格、优化提示词,并通过 GPT Image 2 兼容接口生成图片。

一句话选型结论

按资料字段整理

古诗词生图器是内容创作类的 Agent Skill,安装方式未标注,权限级别未标注,许可证不明确,有可直接复制的安装命令。技能本质是给模型读的说明书,效果取决于模型与客户端;装完先用一个小任务验证,再放进正式流程。

分类内容创作 安装方式未标注 权限级别未标注权限 许可证许可证不明确
累计安装 未标注 本周 未标注
GitHub 星标 未标注 仓库 未标注
下载量 未标注 技能市场下载数
权限级别 未标注权限 安装前请确认风险范围
许可证 许可证不明确 未说明

安装

未标注
在支持 Skills 的客户端里执行,装完问一句「你有哪些技能」验证
skillhub install user_0f026afa/poetry-to-image
  1. 1
    确认客户端支持

    确认你的客户端支持 Skills(Claude、Codex、Cursor 等),并升级到较新的版本。

  2. 2
    拿到技能文件

    执行上面的命令安装;也可以把仓库里的 SKILL.md 手动放进客户端的 skills 目录。

  3. 3
    刷新技能列表

    重新打开客户端,或刷新技能列表,让新放进来的技能被加载。

  4. 4
    验证是否生效

    问一句「你有哪些技能」,确认列表里出现它,再拿一个小任务试跑。

技能档案

内容创作
字段来自技能仓库与公开索引,中文说明由本站标注
作者 / 维护方user_0f026afa仓库 未标注 · 版本 未标注
宿主客户端未标注技能可被这些客户端加载
安装方式未标注安装包 community
权限级别未标注权限安装前请确认风险范围
许可证未说明许可证不明确,商用前请自行确认条款
仓库可用性available技能仓库当前是否可访问
最近核验未标注上游最后核验时间 · 数据更新 1791624972
正文体量7.8 KBSKILL.md 原文大小,越大说明说明书写得越细

典型使用场景

3 类
按分类整理的常见用法,具体以 SKILL.md 为准
写文章与脚本

按主题产出初稿与分镜脚本。

改稿与润色

把草稿改顺、统一语气。

多平台文案改写

同一内容按平台改成长短不同的版本。

选择建议

装之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有可直接复制的安装命令,不用自己翻文档找入口。

先想清楚的地方

  • 没有标注权限级别,实际能做什么要装好后确认。
  • 许可证不明确,商用或二次分发前先到仓库确认条款。
  • 没有做站内核验,内容以作者仓库为准。
  • 技能只是说明书,实际效果取决于模型与客户端,装完先用小任务试一遍。

更适合

  • 需要在内容创作场景里给模型补一套做法的人。
  • 愿意先读一遍 SKILL.md、确认它会让模型做什么的人。

不太适合

  • 对许可证要求很严、必须明确可商用的人。
  • 只想装完就自动变强、不打算看说明书的人。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
技能权限级别安装方式累计安装宿主客户端安装命令
古诗词生图器当前 未标注权限 未标注 未标注 未标注 有
AI 视频生成 调用外部服务 GitHub 仓库 431.3 万 Codex、Claude、Cursor 有
视频帧 联网访问 ClawHub 市场 262.1 万 OpenClaw 有
Markdown 转换器 写入文件 ClawHub 市场 135.3 万 OpenClaw 有
哔哩哔哩全合一 接触敏感数据 ClawHub 市场 64.1 万 OpenClaw 有
Remotion video toolkit 命令行 浏览器控制 ClawHub 市场 63.9 万 OpenClaw 有

SKILL.md 全文

7.8 KB
模型实际读到的说明书原文,来自技能仓库

Poetry to Image

Turn Chinese poetry into a coherent visual concept and a generated image. The deterministic scripts use only the Python standard library; Codex supplies the semantic judgment and final prompt refinement.

Fixed locations

Keep source code and user data separate:

  • This skill: the directory containing this SKILL.md, normally $HOME/.agents/skills/poetry-to-image/.
  • Private configuration: $HOME/poetry-to-image/.env.
  • Generated works: $HOME/poetry-to-image/outputs/<run-id>/.

Honor POETRY_TO_IMAGE_HOME when it is already set. Never store credentials or generated works inside the installed skill directory.

Before running

  1. Resolve the absolute skill directory from this SKILL.md; do not assume the current working directory is the skill directory.
  2. Select a local Python 3.10+ command: prefer python3, then py -3 on Windows, then python.
  3. If Codex requests permission, ask only for:
    - network access to https://poetry-api.1token-store.com and https://edge.1token-store.com;
    - read/write access to $HOME/poetry-to-image/.
  4. Do not ask the user to enable Full access. If a managed policy blocks the home directory, ask the user to attach or open $HOME/poetry-to-image/ as a local folder instead.

Secret configuration

Check configuration with:

<PYTHON> <SKILL_DIR>/scripts/config_manager.py show

If api_key is empty:

  1. Tell the user to apply for a dedicated, revocable key at https://1token-store.com/keys.
  2. Ask the user to send it in the current private conversation.
  3. Do not quote, summarize, log, or repeat the key.
  4. Start this command without putting the key in the command line:
<PYTHON> <SKILL_DIR>/scripts/config_manager.py configure --key-stdin
  1. Send the key only through the host's process-stdin facility. Never interpolate it into a shell command, filename, prompt, JSON output, or image metadata.
  2. Run config_manager.py show and confirm only [REDACTED] is displayed. On Unix, verify .env mode is 0600; the script applies it automatically.
  3. Suggest that the user delete their key-bearing chat message after configuration. Do not require them to edit .env manually.

The default service settings are:

  • Poetry API: https://poetry-api.1token-store.com
  • Image API: https://edge.1token-store.com
  • Image model: gpt-image-2

Default workflow

1. Select a poem and prepare a run

For a theme, line, title, or author:

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py prepare --query "<USER_QUERY>"

For a random complete poem:

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py prepare

The search script expands classical imagery, queries multiple terms, merges and deduplicates results, probes exact titles, locally reranks candidates, and rejects fragments. Never replace this with “take the first API result.”

Read the returned poem_path and draft_prompt_path.

2. Refine the final prompt

Use the draft as a strong starting point. Preserve:

  • the exact poem and author; preserve dynasty only as upstream source metadata, and do not use an unverified dynasty value to force period-specific clothing or architecture;
  • concrete imagery and season/time clues;
  • one coherent scene rather than an inventory of objects;
  • clear foreground, middle ground, background, light, atmosphere, material, and palette;
  • the selected original style's medium and composition;
  • low-text or no-text output, with no watermark, logo, QR code, brand, or real celebrity.

Apply the user's requested style, mood, subject emphasis, and aspect ratio. Save the refined prompt as <run_dir>/prompt.txt. Do not silently replace the poem with a more famous one after the user has approved it.

3. Configure the key if needed

If needs_image_key is true, complete Secret configuration before generation.

4. Generate and display

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py generate --run-dir "<RUN_DIR>"

The command uses prompt.txt when present, otherwise draft_prompt.txt. If the run already contains a verified generated image, it returns that image without calling the provider again. Use --force only when the user explicitly asks to regenerate the same run and accepts another billable request. After success:

  1. Verify image_path exists and is a parseable PNG, JPEG, or WebP. Record its actual pixel dimensions. Never discard or fail a valid paid image because its ratio differs from the requested aspect.
  2. Display or attach the local image directly in the conversation. If aspect_mismatch is true, mention the provider-returned dimensions as a non-blocking note.
  3. Give the poem title, author, selected style, aspect ratio, and run directory.
  4. Do not upload to IMA or any other archive unless the user separately asks for one.

Focused modes

Search only

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py search "<QUERY>" --limit 5

Return ranked poems with match reasons. Do not generate an image.

Random poem only

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py random

Return the normalized poem. Do not generate an image.

Prompt only

Run prepare, refine the draft, save prompt.txt, and return the poem plus prompt. Do not request an image key when the user does not want image generation.

User-provided poem

Normalize the poem to a UTF-8 JSON file containing title, author, dynasty, and a lines array, then run:

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py prepare --poem-file "<POEM_JSON>"

Use source: "user-provided" when no public source is known.

Reuse the previous poem

Use the previous run's poem.json to create a new run without changing the old output:

<PYTHON> <SKILL_DIR>/scripts/poetry_to_image.py prepare --poem-file "<PREVIOUS_RUN>/poem.json" --style "<STYLE>" --aspect <landscape|portrait|square>

Then refine and generate normally. Never overwrite the previous image or metadata.

Output contract

Each successful run is self-contained:

$HOME/poetry-to-image/outputs/<run-id>/
├── image.png | image.jpg | image.webp
├── poem.json
├── poem.md
├── analysis.json
├── draft_prompt.txt
├── prompt.txt
└── metadata.json

metadata.json may contain provider base URL, model, size, and local paths. It must never contain the API key, Authorization header, raw secret-bearing requests, or hidden Codex instructions.

When the user asks to package or share results, package only the selected run directory or outputs/; never package the parent $HOME/poetry-to-image/ because it contains .env.

Error handling

  • Poetry API unavailable: report the endpoint and error; do not invent a poem or API response.
  • No quality search result: broaden terms once, show candidates, and ask the user to choose if still ambiguous.
  • Missing key: follow Secret configuration.
  • HTTP 401/403: say the key is invalid or unauthorized without displaying it.
  • HTTP 402/quota error: tell the user to check balance or key limits.
  • HTTP 429: wait briefly and retry once; do not loop.
  • HTTP 5xx/network timeout: retry once, preserve the prepared run, and report the blocker.
  • Invalid image bytes: do not save them with an image extension and do not claim generation succeeded.

References

Read only when needed:

  • references/poetry-api.md — endpoint and response contract.
  • references/security-and-storage.md — sandbox, secret, and sharing rules.
  • assets/prompt-styles.json — original compact visual-style library.
  • assets/motif-lexicon.json — imagery extraction rules.

常见问题

5 问
答案由本站依据条目字段整理
古诗词生图器怎么安装?

在支持 Skills 的客户端里执行:skillhub install user_0f026afa/poetry-to-image。安装方式为未标注。装完刷新技能列表或重开客户端。

古诗词生图器支持哪些客户端?

资料没有标注宿主客户端,但只要客户端支持 Skills(Claude、Codex、Cursor 等),一般都能加载。

古诗词生图器权限高吗,安全吗?

权限级别为未标注权限。安装前先看它会读取或写入哪些内容。

古诗词生图器收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于技能本身和所用的模型或客户端,正式使用前按作者说明确认。

古诗词生图器用什么许可证?

许可证标注为 未说明(许可证不明确)。商用或二次分发前请到仓库确认条款原文。

热门专题

按场景挑技能
全部专题 →

技能内容以作者仓库为准;分类、权限级别与许可证状态的中文说明为中文整理,最近核验 未标注。