Frontier-grade Claude models with agentic workflows, strong coding, and enterprise guardrails delivered via developer console, web app, and cloud partners.
AgentLens
0Popularity
About AgentLens
AgentLens is a Chrome extension designed to expose hidden instructions within web pages and PDFs that may be invisible to human users but accessible to AI agents. It scans content for suspicious signals such as instruction overrides, credential requests, data-exfiltration attempts, tool-use commands, concealment tactics, and obfuscation patterns. The tool provides scan reports that detail each flagged finding, including explanations of why it was detected, how it was concealed, and which extraction profiles might encounter it. Users can review the source of hidden instructions, strip suspicious content before sharing with AI tools, or allow specific content to be read. It supports scanning PDF text layers, annotations, metadata, clipped content, and off-page content without requiring document uploads. Optional live protection continuously monitors pages, while local scan history and customizable deterministic rules enhance control over AI agent access to content.
Key features
- Scan web pages and supported PDFs for hidden instructions
- Reveal and highlight suspicious signals with source attribution
- Strip suspicious content before sharing with AI tools
- Review detailed scan reports with explanations and rule matches
- Enable live protection for continuous monitoring
- Maintain local scan history without cloud storage
- Customize deterministic rules for detection
- Export bounded reports on demand
Use cases
- Inspecting web pages for AI agent vulnerabilities before sharing
- Reviewing PDFs for hidden instructions that may leak data
- Verifying content safety for AI tool integration
Pros
- Detects hidden instructions and obfuscation patterns in web pages and PDFs
- Provides detailed explanations for flagged findings and their concealment methods
- Allows stripping suspicious content before sharing with AI tools
- Supports on-device processing without document uploads
- Offers optional live protection and local scan history
Cons
- Chrome extension only, no support for other browsers
- Version 0.1.0 indicates early-stage development
- No API or integration options mentioned
Frequently asked questions about AgentLens
What does AgentLens do?
AgentLens is a Chrome extension that reveals hidden instructions within web pages and PDFs that are typically invisible to human users but may be accessible to AI agents. It scans for suspicious signals such as instruction overrides, credential requests, and obfuscation patterns, providing detailed scan reports to help users understand and control what AI tools can read.
Who should use AgentLens?
AgentLens is designed for developers, security professionals, and users who work with AI tools and want to ensure their content is not inadvertently exposing sensitive instructions or data to AI agents. It is particularly useful for those handling PDFs or web content that may contain hidden prompts or commands.
How does AgentLens work?
AgentLens scans the current page or supported PDFs to detect hidden instructions, credential requests, and obfuscation patterns. It provides a scan report explaining each flagged finding, its concealment method, and which AI extraction profiles might encounter it. Users can then choose to reveal, strip, or allow the content before sharing it with an AI tool.
Does AgentLens require uploading documents to a server?
No, AgentLens performs scans locally on your device. It can analyze PDF text layers, annotations, metadata, clipped content, and off-page content without requiring document uploads, ensuring privacy and security.
Can I customize AgentLens to fit my needs?
Yes, AgentLens offers customizable deterministic rules, local scan history, and the ability to export bounded reports. Users can also enable optional live protection to continuously monitor pages for hidden instructions.
How do I get started with AgentLens?
To get started, add the AgentLens Chrome extension to your browser from the Chrome Web Store. Once installed, you can scan the current page or a supported PDF directly within your browser and review the scan reports to take appropriate action on any detected hidden instructions.
