- Home
- Skills
- Agents & Orchestration
- syncralis web agent
syncralis web agent
Secure MCP web agent with trust scoring, SSRF protection, and gated browser actions to prevent prompt injection.
Free
Works with the AI Agents you already use
syncralis web agent
Example session with this skill installed
Search for the latest whitepaper on the Syncralis homepage and download it if it looks official.
- Read your context and instructions
- Compiled the syncralis web agent
- Generated the document
I have performed a search and located the official site. The page has been analyzed for downloadable content.
Web Research & Download Status
| Source | URL | Trust Score | Finding |
|---|---|---|---|
| Syncralis Official | https://www.syncralis.io | 92 (High) | Located whitepaper_v2.1.pdf |
| Verification | Verified | Signature: OK | Type: application/pdf |
The download for whitepaper_v2.1.pdf was auto-approved because the domain syncralis.io exceeds the trust threshold (80) and passed all security signature checks.
File Details:
- Path:
/tmp/downloads/whitepaper_v2.1.pdf - Size: 2.4 MB
- SHA-256:
e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 - Reputation: Checked via VirusTotal (Clean)
Next steps
- Ask me to read the first 4000 characters of the PDF if you want a summary.
- Review the SHA-256 hash against the organization's public security page.
- Check the audit logs at the configured
AUDIT_LOG_PATHfor a full trace of this session.
1.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Modern LLMs lack live web access and can be easily manipulated by prompt injection hidden on websites. This skill provides a managed browser interface via the syncralis-web-agent MCP server, adding a security layer between the model and untrusted web content.
What it does
- Gated actions enforces a confirmation protocol for risky operations like clicks and downloads when site trust scores are low.
- Trust scoring evaluates domains based on HTTPS status, TLD reputation, typosquatting detection, and curated allowlists.
- SSRF protection blocks requests to local, private, and cloud metadata IP ranges to prevent internal network scanning.
- Injection detection flags hidden text and potential prompt injection signals found within page content.
- Verified downloads performs signature checks, magic-byte verification, and optional VirusTotal reputation lookups on files.
How it works
- Use
research_queryorweb_searchto find sources, specifying brands to improve trust heuristics. - Load specific pages with
open_pageto inspect text previews and identified clickable or downloadable elements. - Initiate a
download_fileorclick_on_pagecall. - If trust is below the threshold, capture the
confirmationIdand present the security reasons to the user. - Use
confirm_actiononly after the user provides explicit manual approval in the chat.
Frameworks & tools
Requires the syncralis-web-agent MCP server running on Node.js 18+. It utilizes Playwright (Chromium) for headless browsing and integrates with Tavily or Brave APIs for search results and VirusTotal for file reputation.
Why this beats prompting it yourself
Standard browsing tools often follow instructions found inside page text, leading to accidental prompt injection. This skill implements a strict state machine that separates user instructions from web data, ensuring the agent never confirms a transaction or starts a download without a manual circuit-breaker.
Use cases
- Researching organizations while surfacing potential typosquatting or phishing warnings.
- Safely downloading verified documents from medium-trust portals.
- Gathering official social media updates using authenticated API routes.
- Analyzing sites for hidden text or malicious injection signals before interacting.
Known limitations
The tool does not bypass CAPTCHAs, logins, or anti-bot walls. Text previews are capped at 4000 characters per page. Staged actions are kept in memory and expire after 15 minutes.
No CAPTCHA or paywall bypass. 4000 character text cap per page. In-memory confirmation IDs expire after 15 minutes. Node 18+ and Chromium libraries required.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 11 days ago
- Free with an Agensi account
Requires
Permissions
Allowed Hosts