instar@1.3.1415
Coherence infrastructure for self-evolving AI agents — on the Claude Code or Codex subscription you already have.
AI Security Review
scanned 1d ago · by lpm-firewall-aiThe runtime startup path silently registers a first-party MCP server and pre-accepts its trust dialog, leaving a concrete agent configuration risk. The install hook is a SQLite repair script, and the inspected credential code uses local Keychain operations, so the evidence supports a warning rather than a malicious block.
Decision evidence
public snapshot- Server startup calls Threadline bootstrap, which invokes MCP registration.
- Registration adds the Threadline server to the project entry in the user's Claude configuration and marks its trust dialog accepted without a consent check in that path.
- The registration also writes project-level MCP configuration; its command launches this package's Threadline MCP entrypoint.
- Other inspected agent setup writes occur in init, setup, or migration paths, while worktree metadata and relay-hook writes have separate product purposes.
- The inspected credential snippets read and write Claude credentials through the local macOS Keychain.
- Quota events pass notification text to an injected sender and queue failed sends; this is operational notification behavior.
- The package postinstall runs a SQLite repair script; the inspected agent configuration behavior is reached through server startup or CLI paths.
- The inspected credential paths show local Keychain handling, not transfer to a remote endpoint.
Source & flagged code
18 flagged · loading sourcePackage defines install-time lifecycle scripts.
package.jsonView on unpkgPackage source references child process execution.
dist/threadline/PipeSessionSpawner.jsView on unpkg · L15Package source references shell execution.
Findings
Affected versions and remediation
This report applies to instar@1.3.1415.
See all versions of instar for other recorded verdicts.
- Review the evidence and your use of instar@1.3.1415 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.