Skills web-search
📦

web-search

Content revision r1 Medium Risk ⚙️ External commands🌐 Network access

Search and Extract Web Research

Research assistants need current sources without manual browsing. This skill runs Tavily and Exa apps through belt for search, extraction, and answers.

Supports: Claude Codex Code(CC)
⚠️ 50 Poor

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request
Review the Skillstore skill "web-search" from https://skillstore.io/skills/halt-catch-fire-web-search.md and its manifest at https://skillstore.io/api/skills/halt-catch-fire-web-search/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/halt-catch-fire-web-search.md Signed manifest GET /api/skills/halt-catch-fire-web-search/manifest Signed lockfile GET /api/skills/halt-catch-fire-web-search/lockfile

Test it

Using "web-search". Search for privacy risks in browser extensions.

Expected outcome:

A concise answer with source links, brief context for each source, and topics that need follow-up review.

Using "web-search". Extract two policy pages and summarize the differences.

Expected outcome:

Clean page text, a comparison summary, and a list of claims that require source verification.

Using "web-search". Find recent research on retrieval augmented generation evaluation.

Expected outcome:

Relevant links, provider-specific search notes, and a short briefing organized by evaluation theme.

Security Audit

Medium Risk
v3 • 7/5/2026 Open versioned report

The skill is a legitimate web search and extraction wrapper, but it depends on external CLI execution and third-party services. Most static backtick findings are Markdown false positives, while the real concerns are belt command execution, wildcard belt permissions, remote installs, and disclosure of user queries or extracted content.

1
Files scanned
153
Lines analyzed
12
Review items
0
False positives ignored

Confirmed security concerns (2)

Medium
SKILL.md:4
Broad Belt CLI Permission
The front matter allows `Bash(belt *)`, which permits any belt subcommand, not only the listed search and extraction apps. A prompt could request unintended belt operations within that wildcard scope.
Line 4 explicitly grants wildcard Bash access to the belt CLI. The file later documents multiple belt subcommands, confirming the permission is operational context rather than incidental text.
Medium
SKILL.md:49-61SKILL.md:68-91SKILL.md:99-121
Third-Party Query and Content Disclosure
The documented workflows send search queries, URLs, extracted page content, and prompts to Tavily, Exa, inference.sh, and OpenRouter. This is expected functionality, but users could expose sensitive research data.
The examples show belt app calls for search, extraction, and OpenRouter Claude models with user-supplied queries, URLs, and prompts. That provides direct evidence of third-party data transfer.
Capability review items (12)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
SKILL.md:7
Ruby/shell backtick execution
> **Install the belt CLI skill:** `npx skills add belt-sh/cli`
The line tells users to run `npx skills add belt-sh/cli`, which installs code from an external package source. That is a real supply-chain exposure, even though it is documented setup.
Medium
SKILL.md:19-24
Ruby/shell backtick execution
```bash
The fenced example runs `belt login` and `belt app run tavily/search-assistant`, invoking an external CLI that sends the query to a remote service. The command is intentional but still creates command execution and network exposure.
Medium
SKILL.md:48-52
Ruby/shell backtick execution
```bash
The example runs `belt app run tavily/search-assistant` with a user query. This is legitimate functionality, but it executes an external CLI and shares query data with a remote provider.
Medium
SKILL.md:58-62
Ruby/shell backtick execution
```bash
The example runs `belt app run tavily/extract` with URL input, causing external retrieval and content processing. User-provided URLs are sent through the CLI, so the external command risk is real.
Medium
SKILL.md:68-72
Ruby/shell backtick execution
```bash
The example invokes `belt app run exa/search`, which executes a local CLI and sends the query to Exa through the platform. This is intended behavior but still an external command path.
Medium
SKILL.md:78-82
Ruby/shell backtick execution
```bash
The example invokes `belt app run exa/answer` with a user question. It is a real external CLI call that can disclose submitted questions to a remote service.
Medium
SKILL.md:88-92
Ruby/shell backtick execution
```bash
The example invokes `belt app run exa/extract` with a URL. It intentionally uses an external command and remote extraction service, which carries network and data-sharing risk.
Medium
SKILL.md:98-108
Ruby/shell backtick execution
```bash
The workflow runs belt search and an OpenRouter Claude model, writes search_results.json, and sends retrieved context to another service. This is useful, but it expands command execution and data-sharing exposure.
Medium
SKILL.md:112-122
Ruby/shell backtick execution
```bash
The workflow runs Tavily extraction, writes content.json, and sends extracted content to an OpenRouter Claude model. This is a real external command chain with possible sensitive content disclosure.
Medium
SKILL.md:134-143
Ruby/shell backtick execution
```bash
The code block recommends several `npx skills add` commands, which install additional skills from external sources. That creates a real supply-chain risk if copied or automated.
Medium
SKILL.md:143-145
Ruby/shell backtick execution
```
The location includes inline `belt app store`, which invokes the external CLI to browse remote apps. The risk is limited, but it is still a real external command path.
Low
SKILL.md:13
Hardcoded URL
![Web Search & Extraction](https://cloud.inference.sh/app/files/u/4mg21r6ta37mpaz6ktzwtt8krr/01kgndq
The Markdown image loads from cloud.inference.sh when rendered, creating a third-party network request and possible tracking surface. It is not malicious, but the hardcoded external media dependency is real.

Risk Factors

⚙️ External commands (25)
SKILL.md:7 SKILL.md:17 SKILL.md:19-24 SKILL.md:24-33 SKILL.md:33-34 SKILL.md:34-40 SKILL.md:40-41 SKILL.md:41-42 SKILL.md:42-48 SKILL.md:48-52 SKILL.md:52-58 SKILL.md:58-62 SKILL.md:62-68 SKILL.md:68-72 SKILL.md:72-78 SKILL.md:78-82 SKILL.md:82-88 SKILL.md:88-92 SKILL.md:92-98 SKILL.md:98-108 SKILL.md:108-112 SKILL.md:112-122 SKILL.md:122-134 SKILL.md:134-143 SKILL.md:143-145
🌐 Network access (9)
SKILL.md:11 SKILL.md:13 SKILL.md:17 SKILL.md:60 SKILL.md:90 SKILL.md:115 SKILL.md:149 SKILL.md:150 SKILL.md:151
Audited by: codex View Audit History →
Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report
Security Assessment

Copy report link

https://skillstore.io/skills/halt-catch-fire-web-search/audits/3?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/halt-catch-fire-web-search/security.svg)](https://skillstore.io/skills/halt-catch-fire-web-search?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/halt-catch-fire-web-search?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/halt-catch-fire-web-search/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/halt-catch-fire-web-search.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>
Academic citations (APA · BibTeX · CFF)

APA citation

halt-catch-fire. (2026). web-search security audit report (audit version 3) [Author version unspecified]. Skillstore. https://skillstore.io/skills/halt-catch-fire-web-search/audits/3

BibTeX citation

@techreport{halt-catch-fire-halt-catch-fire-web-search-2026, author = {halt-catch-fire}, title = {web-search security audit report (audit version 3)}, institution = {Skillstore}, year = {2026}, number = {3}, url = {https://skillstore.io/skills/halt-catch-fire-web-search/audits/3}, note = {Author version unspecified} }

CITATION.cff

cff-version: 1.2.0 message: "If you use this Skill, cite its author and this versioned security audit report." title: "web-search security audit report (audit version 3)" version: "unspecified" type: report authors: - name: "halt-catch-fire" date-released: "2026-07-05" url: "https://skillstore.io/skills/halt-catch-fire-web-search/audits/3" identifiers: - type: other value: "skillstore:halt-catch-fire-web-search:audit:3" description: "Skillstore immutable audit report identifier"

Compare variants

11 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Higher Skillstore usage
qu-skills Recommended

qu-skills-web-search

Skillstore Score 70
Evidence Confidence High
Skillstore usage 54
Updated

2026-08-21

inference-skills

inference-skills-web-search

Skillstore Score 70
Evidence Confidence High
Skillstore usage 12
Updated

2026-08-21

infsh-skills

infsh-skills-web-search

Skillstore Score 68
Evidence Confidence High
Skillstore usage 9
Updated

2026-08-21

skillssh

skillssh-web-search

Skillstore Score 50
Evidence Confidence High
Skillstore usage 30
Updated

2026-08-21

inferen-sh

inferen-sh-web-search

Skillstore Score 50
Evidence Confidence High
Skillstore usage 20
Updated

2026-08-21

halt-catch-fire Current

halt-catch-fire-web-search

Skillstore Score 50
Evidence Confidence Medium
Skillstore usage 5
Updated

2026-08-21

101-skills

101-skills-web-search

Skillstore Score 50
Evidence Confidence Medium
Skillstore usage 5
Updated

2026-08-21

inference-sh-9

inference-sh-9-web-search

Skillstore Score 38
Evidence Confidence Medium
Skillstore usage 24
Updated

2026-08-21

Cain96

cain96-web-search

Skillstore Score 38
Evidence Confidence Medium
Skillstore usage 18
Updated

2026-08-21

inference-sh-skills

inference-sh-skills-web-search

Skillstore Score 38
Evidence Confidence Medium
Skillstore usage 9
Updated

2026-08-21

inference-sh

inference-sh-web-search

Skillstore Score 38
Evidence Confidence Medium
Skillstore usage 8
Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: Medium
55
Architecture
85
Maintainability
85
Content
67
Community
83
Spec Compliance

What You Can Build

Prepare Research Briefs

Collect current web sources and extract article text for concise briefing notes.

Build RAG Context

Gather searchable content from URLs before adding it to retrieval pipelines.

Track Market Signals

Compare public sources and summarize trends using search plus LLM workflows.

Try These Prompts

Basic Web Search 
Search the web for {topic} with Tavily Search Assistant. Return key sources with short summaries.
Extract Article Content 
Extract clean text from these URLs: {urls}. Summarize each source and list images when available.
Compare Search Providers 
Run Tavily Search Assistant and Exa Search for {query}. Compare overlap, unique sources, and source quality.
Build Research Workflow 
Search for {topic}, extract the strongest sources, then draft a briefing with citations and open questions.

Best Practices

  • Confirm sensitive queries can be shared with Tavily, Exa, inference.sh, and OpenRouter before running commands.
  • Use specific questions and known domains when source quality matters.
  • Review returned sources before using answers in reports or agents.

Avoid

  • Do not submit secrets, credentials, or private customer data.
  • Do not treat AI answers as verified facts without source review.
  • Do not run broad extraction jobs without checking target site permissions.

Frequently Asked Questions

Does this skill browse the live web?
Yes. It calls Tavily and Exa through the belt CLI when those services are configured.
What setup is required?
You need the inference.sh belt CLI, an authenticated session, and access to the listed apps.
Does it work with Claude, Codex, and Claude Code?
Yes. It declares support for Claude, Codex, and Claude Code when belt is available.
Can it extract any website?
No. Site access, paywalls, provider limits, and site policies can restrict extraction.
Which services receive my data?
Queries and URLs can be sent to inference.sh, Tavily, Exa, and OpenRouter in the documented workflows.
Are search answers guaranteed accurate?
No. Review sources and verify important claims before using results in decisions.

Developer Details

License

MIT

Skillstore revision

r1

Version notice

The author did not declare a version.

Repository

https://github.com/halt-catch-fire/skills/tree/main/tools/llm/web-search/

Ref

30c73eac2afe762f6aa9c4553158769369d47351

Maintenance freshness

7/18/2026

Usage

3 downloads · 0 views

File structure

📄 SKILL.md