TL;DR
I built AgentShare Agent Readiness, a free Chrome extension designed for the upcoming agent economy. It provides ARS scoring, MCP discovery, and (pending Chrome review in v0.5.0) a local client-side Prompt Injection Scan.
-
Live now (v0.4.0): Agent Readiness Score (ARS),
robots.txt/llms.txtscan, and MCP Connect for Cursor/Claude/VS Code. - Solana Power: It seamlessly complements our AgentShare MCP Server, which empowers AI agents with real-time DeFi intelligence and Solana/Meteora commerce tools.
- Submission Date: July 10, 2026.
What is Agent Readiness?
Agent Readiness means a website or an application stack is fully prepared for autonomous AI clients — such as crawlers, MCP tools, and headless browsers — not just human visitors using Chrome.
The extension evaluates two critical lenses:
| Lens | Core Question |
|---|---|
| Discovery | Can agents easily find your site's policy and intent? (llms.txt, robots.txt, MCP cards) |
| Exposure | What paths are crawlers allowed to hit — and what critical assets does your analytics miss? |
📊 Agent Readiness Score (ARS) is a transparent 0–100 heuristic fully documented at agentshare.dev/docs.
Why This is My Passion Project ❤️🔥
DEV’s Passion Edition asks us to build something inspired by what we truly care about. For me, that obsession is the Agent Economy — not as a passing tech buzzword, but as a core web wiring problem.
- Agents are the new users: Agents will actively read the web to research, analyze, and execute actions.
-
The stakes are financial: Protocols like
x402are moving agents toward holding autonomous budgets and paying for API or content access. - The Risk: When an AI agent possesses a wallet and browsing capabilities, prompt injection stops being a chat novelty and becomes a direct payment risk.
Hidden text on a page (display:none, off-screen positioning, or malicious instruction-like strings) can easily land in an agent’s LLM context window. The agent doesn’t "see" the visual layout like a human; it only reads the serialized text.
This exact obsession led me to build the AgentShare ecosystem (MCP + Solana DeFi integrations) and now, this extension.
What Ships Today vs. What’s Pending
🚀 Live on Chrome Web Store — v0.4.0
- ARS site scan (client-side fetch of public configuration files).
- AI crawler policy breakdown table (
GPTBot,ClaudeBot, etc.). - MCP discovery probe (
server-card.json,.mcp.json). - MCP Connect Tab: Instant config copying for major IDEs (Cursor, Claude Desktop, VS Code).
👉 Install from Chrome Web Store
⏳ Submitted & Awaiting Google Review — v0.5.0
(Submitted July 10, 2026. Store may take 3–5 days to approve due to new scripting permissions).
- Prompt Injection Scan (New): Scans the active tab's DOM instantly when you open the popup, flagging instruction-like patterns or hidden malicious text. 100% client-side only — no data ever leaves your machine.
How the Prompt Injection Scan Works (And What It Does NOT Do)
This is a heuristic, pattern-based scanner — similar in spirit to Web Application Firewall (WAF) string rules or input sanitization lists:
- Catches: High-risk phrases (e.g., "ignore previous instructions", "disregard your system prompt"), and hidden CSS styling containing injection keywords.
-
Skips: Safe zones like
<code>,<pre>,<script>,<style>, and common accessibility classes (sr-only).
🔒 Reality Check for Builders
| What the extension claims | The Reality |
|---|---|
| "Blocks prompt injection everywhere" | No. It only scans one active tab in your browser, entirely on-demand. |
| "AI-powered security analysis" | No. No external LLM is called to interpret the page. It's pure deterministic heuristic matching. |
| "Guarantees safe crypto wallet actions" | No. Wallet safety requires robust tool gating, strict human approval, and protocol-level controls. |
The Solana Connection & The Big Picture
The extension sits perfectly alongside the AgentShare MCP Server. While the extension focuses on DOM hygiene and Discovery, our MCP server handles the data and transactional side on the Solana Blockchain.
By exposing Solana/Meteora DeFi briefs, price feeds, and automated commerce tools directly to AI Agents, we are building the trust and data layer needed for a secure, agent-driven Web3 ecosystem.
Challenge Submission Summary
| Field | Value |
|---|---|
| Project Name | AgentShare Agent Readiness (Chrome Extension) |
| Passion Focus | Agent Economy Infrastructure & Trust Layers |
| Prize Category | Best use of Solana (Integrated via AgentShare MCP) |
| Live Artifact | Chrome Web Store Link |
| Official Docs | agentshare.dev/docs |
FAQ
Q: Is v0.5 available if I install it right now?
A: The Web Store currently serves v0.4.0 until Google approves the update. It will roll out automatically to all users upon approval.
Q: Does the extension log or upload my browsing data?
A: Never. ARS only fetches public metadata, and the Prompt Injection Scan runs locally in your tab via Chrome’s isolated scripting API.
Q: Will it catch 100% of prompt injections?
A: No. Novel obfuscation vectors, images, and post-LLM tool stages are out of scope. Treat its findings as an early warning signal.
Call to Action
- 🛠️ Try the Extension: Chrome Web Store
- 🔍 Scan Your Site: agentshare.dev/scan
- 💬 Feedback: Let me know what broke on your favorite UGC or citation pages in the comments below!
Disclaimer: AgentShare Agent Readiness is a heuristic tool. It does not replace secure agent architecture, strict tool-gating, or protocol-level safeguards.
Top comments (0)