Inspiration
Online opportunities move faster than people can verify them. Financial offers, Web3 launches, giveaways, and promotional claims often rely on confidence and urgency instead of clear evidence.
We wanted to build a simple trust layer that helps people pause before acting. SignalShield was inspired by one question: what if every confident claim had to show its receipts?
What it does
SignalShield lets users submit a claim and up to five public source URLs. It extracts readable source text, finds relevant excerpts, compares evidence across sources, and highlights risk signals or missing information.
The result is a shareable trust report with four possible evidence states:
- Corroborated — multiple sources support the claim
- Single-source — only one source supports the claim
- Conflicting — sources contain opposing evidence
- Unsupported — no matching evidence was found
SignalShield intentionally does not turn an empty evidence list into proof. It labels uncertainty instead of hiding it.
How we built it
The prototype uses a lightweight Node.js HTTP server with a responsive HTML, CSS, and JavaScript interface. The backend validates public URLs, blocks private-network targets, extracts source content, builds an evidence ledger, and compares support across multiple sources.
We also built an MCP server exposing:
inspect_claimcreate_trust_report
This allows other AI agents to use the same evidence-first trust layer instead of inventing their own opaque risk scores.
What we learned
The biggest lesson was that transparency is more useful than a confident-looking score. A source excerpt, URL, and clear explanation help users make better decisions than a mysterious number.
We also learned that small, deterministic building blocks can create a strong product demo: URL safety checks, evidence extraction, source comparison, persistence, shareable reports, and MCP interoperability.
Challenges
The main challenges were handling untrusted URLs safely, preventing fabricated evidence, comparing contradictory source language, and making the results understandable on a mobile screen.
We addressed these by blocking private-network URLs, refusing unsupported claims, showing exact excerpts, labeling source quality, escaping report content, and testing the core behavior with automated tests.
SignalShield is a prototype, not a financial, legal, or security guarantee. Its purpose is to make uncertainty visible before a user takes an irreversible action.
Built With
- api
- css
- evidenceextration
- github
- html
- javascript
- json-rpc
- mcp
- node.js
- render
- rest
- ssrfprotection
- websecurity
