← 返回 Skills 市场
evolinkai

Multi Summarize

作者 EvolinkAI · GitHub ↗ · v1.3.1 · MIT-0
cross-platform ✓ 安全检测通过
198
总下载
0
收藏
0
当前安装
10
版本数
在 OpenClaw 中安装
/install multi-summarize
功能描述
Fast multi-format summarization — paste a URL, drop a file, get the key points. Powered by evolink.ai
使用说明 (SKILL.md)

Summarize

Summarize any content with Claude Opus 4.6 — the most capable text model available.

When to Use

  • User says "summarize this", "what's this about?", "give me the key points"
  • User shares a URL, PDF, video, audio file, or long text
  • User asks for meeting notes, article digest, or video recap
  • User wants a TL;DR of any content

Quick Start

bash scripts/summarize.sh "https://example.com/article"
bash scripts/summarize.sh "/path/to/document.pdf"
bash scripts/summarize.sh "https://youtu.be/dQw4w9WgXcQ"
bash scripts/summarize.sh "/path/to/recording.mp3"

Supported Formats

  • Web URLs — Articles, blogs, documentation
  • PDF — Reports, papers, whitepapers
  • Video — mp4, webm, YouTube links
  • Audio — mp3, wav, m4a, podcasts
  • Text — Markdown, plain text, code files

Example

User: "Summarize this article: https://evolink.ai/blog/deepseek-v4-release-window-prep"

Output:

## TL;DR
DeepSeek V4 has been rumored since January 2026 but still hasn't officially
launched as of March. Multiple predicted windows have passed. The model is
expected to focus on coding with 90% HumanEval scores (unverified).

## Key Takeaways
- Reuters reported in Jan that DeepSeek plans a coding-focused model for Feb 2026
- DeepSeek silently expanded context from 128K to 1M tokens in Feb — likely V4 infra testing
- Unverified benchmarks claim 90% HumanEval, 80%+ SWE-bench Verified
- "V4 Lite" appeared on DeepSeek's site in March but was never officially confirmed
- Multiple launch windows (mid-Feb, Lunar New Year, early March) all passed without release

## Action Items
- Do not migrate production workloads until official announcement
- Monitor DeepSeek's official channels for confirmed specs
- Prepare API integration — EvoLink will support V4 on day one

Configuration

Set your Evolink API key:

export EVOLINK_API_KEY="your-key-here"

Default model: claude-opus-4-6 (no configuration needed).

To use a different model:

export EVOLINK_MODEL="claude-sonnet-4-5-20250929"

Get your API key →

Links

Security

Credentials & Network

EVOLINK_API_KEY is required to call the Evolink API for summarization. Extracted content is sent to api.evolink.ai and discarded after the response is returned. No data is stored. Review Evolink's privacy policy before sending sensitive content.

Required binaries: curl, python3, realpath, file, stat. Optional: yt-dlp, ffmpeg, whisper, pdftotext, markitdown.

File Access Controls

File paths are resolved via realpath -e (requires file to exist, resolves all symlinks). Symlink inputs are explicitly rejected.

The resolved path must fall within SUMMARIZE_SAFE_DIR (default: $HOME/.openclaw/workspace). A trailing-slash comparison prevents prefix-bypass attacks.

Sensitive files are blacklisted by name: .env*, *.key, *.pem, *.p12, *.pfx, id_rsa*, authorized_keys, config.json, .bash_history, .ssh, shadow, passwd.

Tiered size limits: 5MB text / 50MB PDF / 1GB media. MIME validation via file --mime-type.

Temporary Files

Transcripts and subtitles created in /tmp during processing are cleaned up after extraction. PID-based filenames prevent collisions.

Persistence & Privilege

This skill does not modify other skills or system settings. No elevated or persistent privileges are requested.

安全使用建议
This skill is coherent with its stated purpose, but before installing: 1) Confirm you're comfortable sending extracted content (including local files inside SUMMARIZE_SAFE_DIR) to https://api.evolink.ai and review Evolink's privacy policy. 2) Only place files you are willing to share in the configured safe directory (SUMMARIZE_SAFE_DIR). 3) Note the script blacklists some common sensitive filenames but the blacklist is not exhaustive — do not store secrets or private keys in the workspace. 4) Be aware the top-level registry metadata appears to omit required env vars even though the script requires EVOLINK_API_KEY; ensure you set that env var from a trusted source. 5) If you need stronger guarantees, inspect or modify the script to add additional filename/content checks or to run against a self-hosted/approved API endpoint.
功能分析
Type: OpenClaw Skill Name: multi-summarize Version: 1.3.1 The skill provides multi-format summarization (URL, PDF, media) via the Evolink API. The implementation in `scripts/summarize.sh` is notably robust, featuring multiple security layers including path traversal protection via `realpath`, symlink rejection, a comprehensive blacklist for sensitive files (e.g., `.ssh`, `.env`, `shadow`), and proper JSON escaping of user content using Python to prevent injection. All network activity is directed to the stated API endpoint (api.evolink.ai), and the skill includes explicit size limits and MIME validation to prevent resource exhaustion or processing of dangerous file types.
能力评估
Purpose & Capability
The skill claims to summarize URLs, PDFs, audio, and video via Evolink.ai and the script implements exactly that: content extraction + POST to api.evolink.ai using an EVOLINK_API_KEY. Required binaries (curl, python3, realpath, file, stat) and optional extraction tools are appropriate for the stated purpose.
Instruction Scope
Runtime instructions and the script stay within the summarization scope: they extract content, build a prompt, and call the Evolink API. The script will fetch remote content and will send extracted local-file contents to the external API. It enforces a sandbox directory (SUMMARIZE_SAFE_DIR) and blacklists some sensitive filenames, but it will still transmit any allowed file's contents to api.evolink.ai — this is expected but important to notice.
Install Mechanism
No install spec and only a small bash script are provided. Nothing is downloaded or written to disk by an installer step. This is low-risk from an install-mechanism perspective.
Credentials
The script requires an EVOLINK_API_KEY (and optionally EVOLINK_MODEL and SUMMARIZE_SAFE_DIR), which is proportionate for a cloud summarization service. There is a minor inconsistency: top-level registry metadata earlier stated 'Required env vars: none', whereas _meta.json and SKILL.md clearly expect EVOLINK_API_KEY. No unrelated secrets are requested.
Persistence & Privilege
The skill does not request elevated or persistent system privileges, does not set always:true, and does not modify other skills or system settings. Autonomous invocation is allowed (platform default) but not excessive here.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install multi-summarize
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /multi-summarize 触发
  4. 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
v1.3.1
Fix: align SAFE_DIR default across README/SKILL.md/script ($HOME/.openclaw/workspace), add Configuration table with all env vars, add required/optional binaries list
v1.3.0
Security hardening: realpath -e, symlink rejection, trailing-slash prefix check, expanded filename blacklist, temp file cleanup, removed anthropic-version header, added envVars/binaries to metadata, added Security section
v1.2.0
Security: Tiered Quotas and Path Pinning (5MB Text / 50MB PDF / 1GB Video)
v1.1.2
Fix: use temp file for API payload to avoid 'Argument list too long' on large content
v1.1.1
Updated API key signup link with UTM tracking
v1.1.0
Revamped SKILL.md: new description, real blog example, list format, API key link, added Links section
v1.0.3
Updated description: highlight Claude Opus 4.6 as the most capable text model
v1.0.2
Fix: corrected API authentication from x-api-key to Bearer token
v1.0.1
Fix: updated install command to use correct slug multi-summarize
v1.0.0
Initial release: multi-format AI summarization (URLs, PDFs, videos, audio, text) powered by evolink.ai
元数据
Slug multi-summarize
版本 1.3.1
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 10
常见问题

Multi Summarize 是什么?

Fast multi-format summarization — paste a URL, drop a file, get the key points. Powered by evolink.ai. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 198 次。

如何安装 Multi Summarize?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install multi-summarize」即可一键安装,无需额外配置。

Multi Summarize 是免费的吗?

是的,Multi Summarize 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Multi Summarize 支持哪些平台?

Multi Summarize 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 Multi Summarize?

由 EvolinkAI(@evolinkai)开发并维护,当前版本 v1.3.1。

💬 留言讨论