mtech-ai-local-runner@0.1.67
⚠ Under reviewAn answer to a colleague ask now goes first to the session that made the ask, when the asking AI passed its client session id (Work Item e1c08b33). Among the sessions an answer names, the runner takes the one listed first, as long as a person typed in it
AI Security Review
scanned 8d ago · by lpm-firewall-aiThe inspected integration setup is explicitly user-invoked, but the behavior of two shipped native executables remains unresolved. Manual review is needed to assess those binaries; their presence alone does not prove malicious behavior.
Decision evidence
public snapshot- The manifest identifies the package and exposes its CLI; it declares no install lifecycle hook.
- The CLI installs integrations during setup only for apps named with --integrations; the integration function writes the corresponding hooks or plugin.
- The required WebSocket excerpt handles protocol and origin headers; it does not establish suspicious behavior.
- Offline analysis could not identify the format or behavior of either shipped native executable.
- The reviewed integration setup is explicitly user-invoked, and the package code describes installing its own hooks or plugin.
- No concrete attack was established in the inspected source; the native executables remain unresolved.
Source & flagged code
5 flaggedPackage text addresses the security reviewer or scanner and tries to influence the review outcome.
README.mdView on unpkgPackage source references dynamic require/import behavior.
dist/src/adapters/codex/pre-model-gate.jsView on unpkg · L29Findings
Affected versions and remediation
This report applies to mtech-ai-local-runner@0.1.67.
See all versions of mtech-ai-local-runner for other recorded verdicts.
- Review the evidence and your use of mtech-ai-local-runner@0.1.67 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.