pincushion-mcp@1.11.20
Visual feedback as agent work packets. Stakeholders drop pins on your live app; your AI coding agent reads each pin (selector, screenshot, DOM, thread, acceptance criteria) via MCP and ships the fix.
AI Security Review
scanned 2mo ago · by lpm-firewall-aiThe package is not malicious on the inspected source evidence, but its MCP upload tool creates a concrete arbitrary-file exfiltration path. No lifecycle hook or automatic payload execution was found.
Decision evidence
public snapshot- The MCP snapshot tool reads any caller-supplied file path without restricting it to image files or the project.
- It base64-encodes that file and posts the bytes to the configured cloud endpoint.
- The default cloud endpoint is a Pincushion Supabase host, but an environment variable can replace it.
- package.json has no install or postinstall lifecycle hook.
- The simulator command execution is reached only through the explicit mobile-wall subcommand and uses fixed executable names.
- The observed cloud synchronization and telemetry are product-aligned; the arbitrary-path upload is the concrete unsafe capability.
Source & flagged code
11 flaggedA single source file combines environment access, network access, and code or shell execution with blocking evidence.
mobile-wall.mjs#virtual:normalized:round1View on unpkg · L121Findings
Affected versions and remediation
This report applies to pincushion-mcp@1.11.20.
See all versions of pincushion-mcp for other recorded verdicts.
- Review the evidence and your use of pincushion-mcp@1.11.20 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.