← 返回 Skills 市场
nano-banana-pdf-edit
作者
Pratik Singhal
· GitHub ↗
· v1.0.1
765
总下载
2
收藏
0
当前安装
2
版本数
在 OpenClaw 中安装
/install nano-banana-pdf-skill
功能描述
Edit PDF files visually using natural language with the nano-pdf CLI tool, powered by Google's Gemini 3 Pro Image (Nano Banana). Use this skill whenever the...
安全使用建议
This skill appears coherent for visually editing PDFs, but be aware of these practical concerns before installing or using it:
- Data exposure: The tool converts pages to images and sends them (and optionally the PDF text if you enable --use-context) to Google Gemini. Do not use it with sensitive or confidential PDFs unless you accept that data will be sent to Google's API and may be billed.
- API key management: Provide a Gemini API key tied to a project with appropriate billing limits. Consider using a restricted/ephemeral key or project to limit blast radius and cost, and revoke it if you suspect misuse.
- Verify packages/sources: The skill instructs you to pip install nano-pdf and references GitHub repos. Inspect the nano-pdf package source (PyPI/GitHub) and confirm the repository owner before installing to avoid installing a malicious package.
- Test safely: Try the workflow on non-sensitive sample PDFs first to confirm results and costs.
- Use options to limit leakage: Use --no-use-context if you do not want the agent to send the full document text as context, and prefer --disable-google-search if you want to avoid additional external lookups.
If you want a stricter assessment, provide the exact nano-pdf package source (PyPI page or the GitHub repo contents) so I can inspect the code that will be installed and called.
功能分析
Type: OpenClaw Skill
Name: nano-banana-pdf-skill
Version: 1.0.1
The skill is classified as suspicious due to the inherent risk of shell injection and broad execution capabilities. The `SKILL.md` instructs the AI agent to construct and execute `nano-pdf` shell commands based on user input, which presents a vulnerability if user input is not rigorously sanitized or quoted, potentially leading to arbitrary command execution. Additionally, the agent is instructed to install system-level dependencies (`pip install`, `brew install`, `apt-get install`), granting it broad permissions to modify the system, which could be exploited if the agent's command generation is compromised by a malicious prompt.
能力评估
Purpose & Capability
The name, description, and declared requirements (python3 and GEMINI_API_KEY) match the behavior described in SKILL.md: invoking the nano-pdf CLI which uses Gemini Image to edit rendered PDF page images. Minor inconsistency: registry metadata listed source/homepage as unknown/none while SKILL.md/README reference GitHub repos (ps06756/nano-banana-pdf-skill and gavrielc/Nano-PDF). Verify the repository origin before installing.
Instruction Scope
SKILL.md stays within the purpose: it instructs the agent to check dependencies, run nano-pdf edit/add commands, and optionally include document context. Important scope note: the workflow converts PDF pages to images and sends them (and, if enabled, full PDF text via --use-context) to an external model (Google Gemini). That is necessary for the feature but means document content is transmitted outside the host system.
Install Mechanism
This is an instruction-only skill with no install spec or bundled code. It asks users to install nano-pdf via pip and standard OS packages (poppler, tesseract), which is proportionate and expected. Because installation happens via pip/OS package managers, users should validate the nano-pdf package source (PyPI/GitHub) before installing.
Credentials
Only GEMINI_API_KEY is required and declared as primaryEnv, which is appropriate for a skill that calls Google's Gemini API. No unrelated credentials or config paths are requested. Note: the key is described as a paid API key—expect billing for image generation and verify API key project/permissions.
Persistence & Privilege
The skill does not request always: true or any elevated persistence. It is user-invocable and uses the platform default for autonomous invocation; that is expected for skills. It does not request to modify other skills or system-wide configs.
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 nano-banana-pdf-skill - 安装完成后,直接呼叫该 Skill 的名称或使用
/nano-banana-pdf-skill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
v1.0.1
- Added metadata to SKILL.md for environment variables, required binaries, and homepage.
- Updated the "Prerequisites" and "Workflow" sections for improved clarity; removed reference to the dependency check script and now instruct users to manually ensure dependencies are installed.
- Minor rewording and formatting improvements throughout documentation.
- Added .DS_Store file; removed evals/evals.json file.
v1.0.0
Initial release of nano-pdf-edit skill for AI-powered visual PDF editing.
- Enables natural language PDF editing via the nano-pdf CLI tool and Google's Gemini 3 Pro Image.
- Supports fixing typos, updating charts, changing branding/colors, adding slides, and any visual PDF change.
- Automatically checks for required dependencies and guides installation.
- Handles both single and multi-page edits, style reference selection, and context options.
- Includes usage instructions, example commands, and troubleshooting tips.
元数据
常见问题
nano-banana-pdf-edit 是什么?
Edit PDF files visually using natural language with the nano-pdf CLI tool, powered by Google's Gemini 3 Pro Image (Nano Banana). Use this skill whenever the... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 765 次。
如何安装 nano-banana-pdf-edit?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install nano-banana-pdf-skill」即可一键安装,无需额外配置。
nano-banana-pdf-edit 是免费的吗?
是的,nano-banana-pdf-edit 完全免费(开源免费),可自由下载、安装和使用。
nano-banana-pdf-edit 支持哪些平台?
nano-banana-pdf-edit 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 nano-banana-pdf-edit?
由 Pratik Singhal(@ps06756)开发并维护,当前版本 v1.0.1。
推荐 Skills