trace-mcp@3.3.0
Framework-aware code intelligence MCP server — 87 framework integrations, 81 languages, 40–50% fewer tokens on average
AI Security Review
scanned 1mo ago · by lpm-firewall-aiThe package has concrete install-time persistence and remote binary-update behavior, which warrants a warning. The inspected source does not establish malicious exfiltration, a self-dependency chain, or unconsented mutation of a foreign AI-agent control surface.
Decision evidence
public snapshot- The automatic postinstall chain runs two scripts that mutate the host.
- On macOS, postinstall writes and bootstraps a persistent launchd service with RunAtLoad and KeepAlive.
- Postinstall fetches a release asset, verifies only a release-provided checksum, then extracts and replaces an installed application.
- The package includes agent-plugin hooks that can intercept Bash tool use when the user explicitly installs the plugin.
- The manifest has no runtime dependency on trace-mcp itself.
- The app updater checks a SHA-256 value and performs a platform trust check before replacement.
- No credential harvesting or secret exfiltration was identified in the inspected lifecycle scripts.
- The agent hooks are packaged as plugins requiring an explicit plugin-install action, not written by npm postinstall.
Source & flagged code
15 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgInstall-time lifecycle script is not statically allowlisted and needs review.
package.jsonView on unpkgPackage source references child process execution.
scripts/count-advertised-tools.mjsFindings
Affected versions and remediation
This report applies to trace-mcp@3.3.0.
See all versions of trace-mcp for other recorded verdicts.
- Review the evidence and your use of trace-mcp@3.3.0 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.