Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

smart summarizer

v1.0.0

Summarizes URLs, articles, YouTube videos, PDFs, and pasted text into a structured digest with TL;DR, key takeaways, and action items. Use this skill wheneve...

0· 73·0 current·0 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for billyhetech/smart-summarizer-v1.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "smart summarizer" (billyhetech/smart-summarizer-v1) from ClawHub.
Skill page: https://clawhub.ai/billyhetech/smart-summarizer-v1
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install smart-summarizer-v1

ClawHub CLI

Package manager switcher

npx clawhub@latest install smart-summarizer-v1
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The name/description (summarize URLs, YouTube, PDFs, pasted text) aligns with the instructions. Requested capabilities (fetch pages, extract transcripts/PDF text, summarize pasted text) are reasonable for the described purpose. The only persistent artifact (saving summaries) is coherent with the archive feature.
!
Instruction Scope
The SKILL.md tells the agent to fetch arbitrary URLs (via web search or direct fetch), extract YouTube transcripts, and extract PDF text — all expected — but it also instructs the agent to auto-activate when a user shares a URL without explicit instruction. That proactive activation combined with automatic fetching and saving of content could cause network requests and local writes the user didn't explicitly consent to. The archive read/write behaviour (scan ~/.openclaw/summaries) is specified and limited to its own path, but it still implies the skill will both write and later read user data.
Install Mechanism
No install spec and no code files (instruction-only). This is low risk from an install perspective because nothing is fetched or written at install time.
Credentials
The skill requests no environment variables, no credentials, and no config paths beyond its own archive directory. That is proportionate to a summarizer. One note: the instructions mention doing web searches or fetching content; if the runtime implementation uses third‑party search or transcript APIs it may need keys, but none are declared.
Persistence & Privilege
The skill persists summaries to ~/.openclaw/summaries/[YYYY-MM-DD]-[slug].md and will scan that archive on request. Writing to its own per-user directory is normal, but combined with the automatic activation heuristic it means the skill may autonomously fetch, store, and later read user-shared content. The skill is not marked always:true and does not request system-wide config changes.
What to consider before installing
This skill appears to do what it says (summarize content) and has no install or credential requirements, but it does two things you should be aware of: (1) it will automatically activate and fetch any URL a user shares (per the instructions), and (2) it saves summaries to ~/.openclaw/summaries and will read that folder when asked. If you care about privacy or accidental fetching of sensitive links, ask the publisher to: disable or make explicit the automatic activation on link-sharing (require user confirmation), allow opting out of local archival or choosing the archive path, and document whether external search/transcript APIs are used and whether they require any credentials. If you decide to install, ensure your agent's autonomy settings require confirmation before fetching external URLs or performing network I/O for unprompted link shares.

Like a lobster shell, security has layers — review code before you run it.

latestvk97aj2pamd40f1gzas7w5wa72984agpc
73downloads
0stars
1versions
Updated 3w ago
v1.0.0
MIT-0

Smart Summarizer

Purpose

Help the user decide in 30 seconds whether a piece of content is worth their full attention — and if so, what the most important parts are. The summary should be a faithful compression, not a reinterpretation. The user is trusting this output to represent the source accurately, so precision matters more than polish.

Detecting Input Type

Identify what was shared and handle accordingly:

  • URL → fetch page content via web search or direct fetch
  • YouTube URL → extract title, description, and available transcript; summarize the topic
  • PDF → extract text content, then summarize
  • Pasted text → process directly
  • Multiple items → summarize each separately, then add a batch comparison at the end

Summary Format

Use this structure for every summary:

📑 [Title]
━━━━━━━━━━━━━━━━━━
Source: [URL or "pasted text"]
Length: [~X min read / ~X words]
━━━━━━━━━━━━━━━━━━

💡 TL;DR:
[One sentence. The core message — what would you tell a friend?]

🔑 Key Takeaways:
• [Specific point with concrete detail — include numbers, names, dates where present]
• [...]
• [3–5 bullets total]

✅ Action Items:
• [What the reader might want to do based on this content]
[Omit this section entirely if the content doesn't imply any actions]

🤔 Worth reading in full?
[One honest sentence: what you'd gain from the full version vs. this summary]

What Makes a Good Summary

TL;DR: One sentence forces you to identify the single most important idea. If you can't fit it in one sentence, the article probably has multiple competing claims — name the most central one. Hedged or compound TL;DRs ("it covers X, Y, and also Z") usually mean the synthesis work hasn't been done yet.

Key Takeaways: The value here is specificity. "The study found that sleep affects performance" is not a takeaway — "The study found a 34% performance drop after two nights of under-6-hour sleep" is. Include numbers, names, and dates when the source has them.

Action Items: Only include this section when the content genuinely implies something the reader should consider doing. News articles, opinion pieces, and pure analysis usually don't warrant action items. Product reviews, how-tos, and research with clear implications do.

Assessment: Be honest. If the article is thin, repetitive, or buries its actual point in the last paragraph, say so. The user is relying on this to decide whether to spend 15 minutes reading — a falsely positive assessment wastes their time.

Batch Mode

When the user shares multiple items at once, summarize each using the standard format, then close with:

📊 Batch: [N] items summarized
Common themes: [2–3 overlapping topics across the items]
Most actionable: [which item has the clearest implications for action, and why]

Archive

Save each summary to ~/.openclaw/summaries/[YYYY-MM-DD]-[slug].md.

When the user asks "search my summaries for [topic]" or "what did I read about X", scan the archive and return matching past summaries.

Language

Detect the content language automatically and match it in the output. If the source is Chinese, respond in Chinese. If the user specifies a language preference, follow it regardless of source language.

When Things Go Wrong

If a URL can't be fetched, report it directly rather than guessing at the content:

⚠️ Couldn't access [URL]
Reason: [paywalled / 404 / access denied]
Suggestion: Try pasting the article text directly.

Only summarize what's actually in the source. If the content is thin, say so rather than padding the output. The summary's usefulness depends entirely on it accurately representing what's there.

Comments

Loading comments...