← 返回 Skills 市场
343
总下载
0
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install openclaw-paperbanana
功能描述
Generate publication-quality academic diagrams, methodology figures, architecture illustrations, and statistical plots from text descriptions using the Paper...
安全使用建议
This skill appears internally consistent and implements the advertised workflow, but remember: (1) it sends whatever text, CSV/JSON, and images you provide to external LLM/VLM/image APIs — don't use it with sensitive or proprietary data unless your policy allows it; (2) it relies on 'uv' and a PyPI package (paperbanana[all-providers]) — verify the PyPI project and the GitHub repos linked in the SKILL.md/README before installing; (3) the README suggests installing 'uv' via a curl|sh command — review that script before running it; (4) provide a provider API key with appropriate billing/permissions and consider using a dedicated key/account for this skill to limit blast radius. The small documentation inconsistency about 'required env vars' is minor but worth noting.
功能分析
Type: OpenClaw Skill
Name: openclaw-paperbanana
Version: 0.1.1
The skill is designed to generate diagrams and plots using external AI providers (Google Gemini, OpenAI, OpenRouter). It transparently discloses that user-provided data (text, images, CSV/JSON) is sent to these third-party APIs, explicitly warning users not to use it with sensitive data. The Python scripts (`evaluate.py`, `generate.py`, `plot.py`) implement this functionality by reading user input files, making API calls, and saving generated images to `/tmp`. Dependencies are managed securely via `uv`, and there is no evidence of prompt injection attempts in `SKILL.md`, hidden malicious code, unauthorized data exfiltration beyond the stated purpose, persistence mechanisms, or obfuscation. The `curl | sh` instruction for `uv` in `README.md` is for user installation, not executed by the agent.
能力评估
Purpose & Capability
Name/description match the included scripts and README. The skill requires an LLM/VLM provider API key (Gemini/OpenAI/OpenRouter) and the 'uv' binary to run the packaged Python scripts — these are reasonable for an on-demand diagram/plot generation skill. The declared primary credential (GOOGLE_API_KEY) fits the documented auto-detection priority (Gemini → OpenAI → OpenRouter).
Instruction Scope
Runtime instructions and scripts explicitly read user-provided inputs (text files, CSV/JSON, image paths) and send them to external LLM/VLM providers for planning, image generation, and evaluation. Generated images may also be sent back to the provider for VLM-based evaluation. This is documented in SKILL.md and is coherent with the stated purpose, but it means any data you pass (including files you point to) will be transmitted to third-party APIs.
Install Mechanism
There is no registry install spec; the skill relies on 'uv' to create an isolated environment and install the PyPI package 'paperbanana[all-providers]'. Using PyPI for the package is expected. The README suggests installing 'uv' via a curl | sh one-liner (remote install script) — that is common but has the usual remote-install risks; verify the 'uv' install script and the PyPI package/project before running.
Credentials
The skill requests provider API keys (GOOGLE_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY) which are necessary for the LLM/VLM and image-generation work it performs. No unrelated credentials, secrets, or system config paths are requested. Minor metadata mismatch: registry lists 'Required env vars: none' while primaryEnv is set to GOOGLE_API_KEY and SKILL.md says at least one provider key is required — this is a documentation inconsistency but not a functional mismatch.
Persistence & Privilege
The skill is not always-enabled and does not request elevated or persistent system privileges. It writes transient output under /tmp and does not modify other skills or system-wide configs. API keys are read from the environment/config and are not persisted by the skill.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install openclaw-paperbanana - 安装完成后，直接呼叫该 Skill 的名称或使用
/openclaw-paperbanana触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v0.1.1
**Expanded documentation and API key instructions**
- Added explicit homepage, PyPI package, and GitHub links to metadata.
- Added a detailed table listing required API keys, providers, and prerequisites.
- Included a new Privacy & Data Handling section explaining use of LLM/VLM APIs and cautioning against use with sensitive data.
- Added Dependencies & Provenance section for transparency on packages and sources.
- Clarified auto-detection and error behavior if API keys are missing.
- Previous usage instructions and reference material remain intact.
v0.1.0
Initial release of PaperBanana — an academic illustration generator.
- Generate publication-quality academic diagrams, methodology figures, system architecture illustrations, and statistical plots from text.
- Supports plot creation from CSV/JSON data and intent-based chart specification.
- Evaluate generated diagrams against reference images on key academic criteria.
- Refine diagrams iteratively with feedback or by continuing previous runs.
- Multi-provider support: auto-selects Gemini (free), OpenAI, or OpenRouter per config.
- Simple CLI for generation, evaluation, and refinement; auto-installs dependencies in isolation.
元数据
常见问题
Paperbanana 是什么？
Generate publication-quality academic diagrams, methodology figures, architecture illustrations, and statistical plots from text descriptions using the Paper... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 343 次。
如何安装 Paperbanana？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install openclaw-paperbanana」即可一键安装，无需额外配置。
Paperbanana 是免费的吗？
是的，Paperbanana 完全免费（开源免费），可自由下载、安装和使用。
Paperbanana 支持哪些平台？
Paperbanana 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Paperbanana？
由 Bennett（@goatinahat）开发并维护，当前版本 v0.1.1。
推荐 Skills