Skills extract

📦

extract

Name: extract
Author: tavily-ai

Content revision r1 High Risk ⚙️ External commands🌐 Network access📁 Filesystem access🔑 Env variables

Extract Clean Web Content with Tavily

Clean webpage content is hard to collect when pages include navigation, scripts, and formatting noise. This skill extracts readable markdown or text from known URLs through Tavily.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

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

Agent request

Review the Skillstore skill "extract" from https://skillstore.io/skills/tavily-ai-extract.md and its manifest at https://skillstore.io/api/skills/tavily-ai-extract/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

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/tavily-ai-extract.md Signed manifest GET /api/skills/tavily-ai-extract/manifest Signed lockfile GET /api/skills/tavily-ai-extract/lockfile

Test it

Using "extract". Extract a documentation page about authentication.

Expected outcome:

A readable markdown page with the title, sections, examples, and the original source URL.

Using "extract". Extract five product pages and focus on pricing details.

Expected outcome:

A source-by-source summary with relevant pricing text, extracted page content, and any failed URLs.

Using "extract". Extract a JavaScript-heavy dashboard page with images.

Expected outcome:

Advanced extraction results with visible text, available image URLs, and notes about missing content.

Security Audit

High Risk

v4 • 7/7/2026 Open versioned report

The skill performs a legitimate Tavily extraction workflow through an included shell script. Confirmed risks are OAuth token cache reads, environment secret handling, bearer token transmission, and unpinned npx execution. Most Markdown backtick, example URL, and /dev/null findings are false positives, and I found no prompt injection text.

Files scanned

369

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (1)

High

scripts/extract.sh:98

Unpinned Package Execution During OAuth

scripts/extract.sh runs npx -y mcp-remote when no token is found. The package version is not pinned, so authentication can execute newly published npm code.

The command directly invokes npx with automatic yes behavior on a package name and no version pin. This is a clear supply-chain execution risk during normal authentication.

Capability review items (18)

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

High

scripts/extract.sh:50

Shell command substitution

token=$(jq -r '.access_token // empty' "$token_file" 2>/dev/null)

The command substitution reads access_token values from token cache files. Those credentials are later exported and used for Tavily authentication.

High

scripts/extract.sh:67

Shell command substitution

token=$(get_mcp_token) || true

This command substitution invokes get_mcp_token, which searches the local MCP auth cache for bearer tokens. It is intended behavior, but it is credential discovery.

High

scripts/extract.sh:107

Shell command substitution

token=$(get_mcp_token) || true

This command substitution repeats token cache discovery during OAuth polling. It can load a local bearer token into the skill process.

High

scripts/extract.sh:151-157

Shell command substitution

RESPONSE=$(curl -s --request POST \

The command substitution executes curl and captures a response from the Tavily MCP endpoint. It sends user-provided extraction arguments and an Authorization header to a third party.

High

scripts/extract.sh:98

Hardcoded URL

npx -y mcp-remote https://mcp.tavily.com/mcp </dev/null >/dev/null 2>&1 &

The hardcoded Tavily MCP URL is used by the OAuth helper during authentication. This creates live third-party network access before extraction runs.

High

scripts/extract.sh:152

Hardcoded URL

--url "https://mcp.tavily.com/mcp" \

The script posts extraction requests to the hardcoded Tavily MCP endpoint. It sends requested URLs, options, and a bearer token outside the local environment.

High

scripts/extract.sh:45

Hidden file access

MCP_AUTH_DIR="$HOME/.mcp-auth"

The script explicitly targets $HOME/.mcp-auth, a hidden auth cache under the user home directory. Reading this directory can expose stored OAuth tokens.

High

scripts/extract.sh:50

Standard device file access

token=$(jq -r '.access_token // empty' "$token_file" 2>/dev/null)

The jq command reads access_token from token files discovered under the MCP auth cache. This is real credential file access, not only device redirection.

High

scripts/extract.sh:60

Standard device file access

done < <(find "$MCP_AUTH_DIR" -name "*_tokens.json" 2>/dev/null)

The find command recursively scans the hidden MCP auth directory for token files. This broad search can discover local credentials available to the process.

High

scripts/extract.sh:66

Generic API/secret keys

if [ -z "$TAVILY_API_KEY" ]; then

The script checks the TAVILY_API_KEY environment variable as part of authentication flow. This is intended, but it is real access to a secret-bearing environment value.

High

scripts/extract.sh:69

Generic API/secret keys

export TAVILY_API_KEY="$token"

The script exports a discovered OAuth token as TAVILY_API_KEY. This places a bearer credential into the process environment for later network use.

High

scripts/extract.sh:94

Generic API/secret keys

if [ -z "$TAVILY_API_KEY" ]; then

The script checks the TAVILY_API_KEY environment variable as part of authentication flow. This is intended, but it is real access to a secret-bearing environment value.

High

scripts/extract.sh:109

Generic API/secret keys

export TAVILY_API_KEY="$token"

The script exports a discovered OAuth token as TAVILY_API_KEY. This places a bearer credential into the process environment for later network use.

High

scripts/extract.sh:120

Generic API/secret keys

if [ -z "$TAVILY_API_KEY" ]; then

The script checks the TAVILY_API_KEY environment variable as part of authentication flow. This is intended, but it is real access to a secret-bearing environment value.

High

scripts/extract.sh:153

Generic API/secret keys

--header "Authorization: Bearer $TAVILY_API_KEY" \

The Authorization header sends TAVILY_API_KEY as a bearer token to Tavily. This is intended API authentication, but it is confirmed secret transmission.

High

SKILL.md:13

Hidden file in home directory

1. Check for existing tokens in `~/.mcp-auth/`

The documentation states that the skill checks ~/.mcp-auth for existing tokens. This documents hidden home-directory credential access that the script also performs.

High

SKILL.md:13

Hidden file access

1. Check for existing tokens in `~/.mcp-auth/`

The documentation states that the skill checks ~/.mcp-auth for existing tokens. This documents hidden home-directory credential access that the script also performs.

Medium

scripts/extract.sh:1

Unix shell invocation

#!/bin/bash

The skill ships an executable shell script that runs local commands, network tools, and auth-token discovery. Script execution is intended, but it expands the skill trust boundary.

Risk Factors

⚙️ External commands (52)

🌐 Network access (32)

📁 Filesystem access (16)

scripts/extract.sh:45 scripts/extract.sh:17 scripts/extract.sh:26 scripts/extract.sh:32 scripts/extract.sh:50 scripts/extract.sh:60 scripts/extract.sh:98 scripts/extract.sh:115 scripts/extract.sh:116 scripts/extract.sh:128 scripts/extract.sh:134 scripts/extract.sh:163 SKILL.md:13 SKILL.md:20 SKILL.md:13 SKILL.md:20

🔑 Env variables (16)

scripts/extract.sh:65 scripts/extract.sh:66 scripts/extract.sh:69 scripts/extract.sh:94 scripts/extract.sh:109 scripts/extract.sh:120 scripts/extract.sh:123 scripts/extract.sh:153 SKILL.md:24 SKILL.md:57 SKILL.md:69 SKILL.md:93 SKILL.md:137 SKILL.md:150 SKILL.md:167 SKILL.md:181

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

Copy report link

https://skillstore.io/skills/tavily-ai-extract/audits/4?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/tavily-ai-extract/security.svg)](https://skillstore.io/skills/tavily-ai-extract?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/tavily-ai-extract?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/tavily-ai-extract/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/tavily-ai-extract.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

tavily-ai. (2026). extract security audit report (audit version 4) [Author version unspecified]. Skillstore. https://skillstore.io/skills/tavily-ai-extract/audits/4

BibTeX citation

@techreport{tavily-ai-tavily-ai-extract-2026,
  author = {tavily-ai},
  title = {extract security audit report (audit version 4)},
  institution = {Skillstore},
  year = {2026},
  number = {4},
  url = {https://skillstore.io/skills/tavily-ai-extract/audits/4},
  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: "extract security audit report (audit version 4)"
version: "unspecified"
type: report
authors:
  - name: "tavily-ai"
date-released: "2026-07-07"
url: "https://skillstore.io/skills/tavily-ai-extract/audits/4"
identifiers:
  - type: other
    value: "skillstore:tavily-ai-extract:audit:4"
    description: "Skillstore immutable audit report identifier"

Compare variants

3 installable variants

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

Why this variant is first

Highest Skillstore Score

Author / source Skillstore Score Evidence Confidence Skillstore usage Updated

pbakaus Recommended

pbakaus-extract

Skillstore Score 79

Evidence Confidence High

Skillstore usage 12

Updated

2026-08-21

tavily-ai Current

tavily-ai-extract

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 10

Updated

2026-08-21

actionbook

actionbook-extract

Skillstore Score 38

Evidence Confidence Medium

Skillstore usage 2

Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

Maintainability

Content

Community

Spec Compliance

What You Can Build

Collect Research Sources

Extract readable content from known articles before summarizing or comparing them.

Review Documentation Pages

Pull clean sections from API docs for implementation planning or support answers.

Monitor Public Pages

Extract content from public product pages for audits, comparisons, and update tracking.

Try These Prompts

Extract One Page

Use the extract skill to extract markdown from this URL: [URL]. Return headings, key points, and the source URL.

Extract Several Sources

Extract content from these URLs: [URL list]. Return one section per source and note any failed pages.

Focus By Topic

Extract these documentation pages with query focus: [topic]. Keep only the most relevant chunks and cite each source URL.

Use Advanced Extraction

Use advanced extraction for these JavaScript-heavy pages. Include image URLs when available and summarize missing or failed content.

Best Practices

Extract only URLs that you are allowed to share with Tavily.
Use query focus when you need specific sections from long pages.
Start with basic extraction, then use advanced mode for missing content.

Avoid

Do not send private, paid, or confidential URLs without approval.
Do not treat extracted content as verified facts without source review.
Do not batch more URLs than the API limit supports.

Frequently Asked Questions

Does this skill require a Tavily account?

Yes. You need an existing Tavily account and a valid OAuth token or API key.

Does it search the web?

No. It extracts content from URLs that you already provide.

What data is sent to Tavily?

The requested URLs, extraction options, and authentication token are sent to Tavily.

Can it handle JavaScript-rendered pages?

Yes. Use advanced extraction for pages where basic extraction misses content.

Where does authentication come from?

The script reads the MCP auth cache or uses the TAVILY_API_KEY environment variable.

Is it safe for private pages?

Use it only when your policy allows sharing those URLs and page data with Tavily.

Developer Details

Author

tavily-ai

License

MIT

Skillstore revision

Version notice

The author did not declare a version.

Repository

https://github.com/tavily-ai/skills/tree/main/skills/tavily/extract/

Ref

36e07d5e13068e5be64447e8f20b427cf2cbd21a

Maintenance freshness

7/18/2026

Usage

6 downloads · 248 views

File structure

📁 scripts/

📄 extract.sh

📄 SKILL.md