@lightupai/polaris@0.0.137
Record and stream Claude Code sessions to Slack. Like Gong for AI coding — capture every prompt, response, and tool call. Multiplayer collaboration across AI coding agents.
AI Security Review
scanned 3d ago · by lpm-firewall-aiThe flagged Slack credential is an application-configured bot token used for signup notifications, while the Google token goes to Google’s own userinfo endpoint. Session capture is the package’s stated purpose, and no install lifecycle hook or unrelated credential destination was found.
Decision evidence
public snapshot- The manifest declares no install lifecycle hook; its only npm script is `test`.
- The README describes recording Claude Code prompts, responses, and tool calls and streaming them to Slack as the package’s purpose.
- Installed capture hooks forward agent events to a local loopback daemon; transcript capture matches the documented session-recording function.
- Signup notices use the configured `SIGNUP_SLACK_BOT_TOKEN` to post signup details to Slack; the function returns without sending if the token is absent.
- The Google access token is sent to Google’s userinfo endpoint as part of OAuth login.
- CLI installation is dispatched by the explicit `install` command; the update check fetches this package’s latest version from npm.
Source & flagged code
6 flaggedPackage source references child process execution.
scripts/perf-audit.tsView on unpkg · L14A single source file combines environment access, network access, and code or shell execution; review context before blocking.
scripts/perf-audit.tsView on unpkg · L33Findings
Affected versions and remediation
This report applies to @lightupai/polaris@0.0.137.
See all versions of @lightupai/polaris for other recorded verdicts.
- Review the evidence and your use of @lightupai/polaris@0.0.137 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.