integram-mcp@0.7.110
MCP server for Integram — connect AI assistants to your workspace via Model Context Protocol
AI Security Review
scanned 4d ago · by lpm-firewall-aiThe runtime updater can change a containing consumer worktree without a confirmation step, which warrants a warning. No install-time hook or concrete malicious payload was established, and the flagged account-action locations do not perform account changes.
Decision evidence
public snapshot- The manifest exposes a command-line entrypoint and has only a prepublish check, with no install lifecycle hook.
- Launching the command runs an updater immediately and schedules it to run hourly.
- The updater checks for a Git worktree from the package directory and runs a fast-forward pull there.
- If that pull changes the commit, the updater runs npm install in the package directory and exits shortly afterward.
- The flagged account-action locations resolve to update-check cache/startup logic and fatal-error handling, not authenticated account changes.
- The executable guard limits this updater path to running the command-line program.
- The manifest contains no preinstall, install, or postinstall hook; the behavior requires a user or host to launch the command.
- The pull uses the Git remote configured for the detected worktree, rather than a fixed publisher endpoint.
Source & flagged code
3 flagged · loading sourceThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
index.jsView on unpkgPackage source invokes a package manager install command at runtime.
index.jsView on unpkg · L3139Findings
Affected versions and remediation
This report applies to integram-mcp@0.7.110.
See all versions of integram-mcp for other recorded verdicts.
- Review the evidence and your use of integram-mcp@0.7.110 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.