Chaintrap Agent Shield

Marketplace · GitHub · Privacy

Workspace visibility for packages introduced by AI coding agents (Cursor, Claude Code, GitHub Copilot). Chaintrap inventories npm and PyPI pins in the open folder and in MCP server configs, classifies each as malicious or vulnerable, and lets the operator remove the pin from the editor.

Chaintrap activity bar: Dependencies and MCP servers grouped as malicious, vulnerable, or unpinned

Coverage

Dependencies: workspace manifests and lockfiles. With a lockfile, locked transitives are included.
MCP servers: the package inferred from each server entry in mcp.json (for example npx or pip). Listed separately from project dependencies so the operator sees which server pulled the pin.
Unpinned MCP servers: config has a package name and no version, so the pin cannot be checked.

Findings are malicious (known-bad or malware advisory) or vulnerable (CVE or GHSA). High and critical findings require an acknowledgement in the editor.

Confirm removal of one MCP server from mcp.json

Confirm removal of a malicious PyPI pin from the manifest

Malicious MCP servers are removed from that config by server id. Malicious npm and PyPI pins are removed from package.json or requirements.txt.

Operation

Install pri17m.chaintrap-agent-shield.
Open a workspace folder. Status bar: Chaintrap: scanning workspace…
Open the Chaintrap activity bar: Dependencies or MCP servers.

No API key is required for this scan. The extension reads dependency and MCP files in open folders. It does not execute workspace code and does not upload source. See PRIVACY.md.

Commands

Command	Action
Chaintrap: Rescan workspace baseline	Run the workspace inventory again
Chaintrap: Review agent changes since last session	List findings that appeared after this window opened
Chaintrap: Acknowledge high/critical finding	Record that a serious finding was reviewed
Chaintrap: Uninstall malicious package	Remove a malware pin or that MCP server from the config
Chaintrap: Open finding location	Open the file that pinned the package

MIT. github.com/pri17m/chaintrap-agent-shield

Chaintrap Agent Shield — AI-Coding Security

Debarshi Ghosh

Chaintrap Agent Shield

Coverage

Operation

Commands