@bongos/core@1.20.46
Cloud Bongos — the AI-first build platform core (GDS + platform surfaces + module system), installed as a versioned dependency (ADR 0108).
AI Security Review
scanned 5d ago · by lpm-firewall-aiWarn because the onboarding feature sends compacted session text to an external model provider while setup says raw transcripts stay local. Consent gates the flow, but the transfer is not clearly disclosed; source inspection does not establish malicious intent.
Decision evidence
public snapshot- package.json shows the package entrypoints and no install lifecycle hook.
- Setup asks to analyze onboarding logs, accepts an empty answer as consent, and says raw transcripts stay on the machine.
- The onboarding pipeline checks the saved consent window, stages completed session transcripts locally, and reads them for analysis.
- The analyzer includes compacted builder and assistant text in a prompt and pipes it to the Claude CLI Anthropic provider.
- After model analysis, the platform inbox request contains the derived title and body.
- The other flagged runtime locations perform route normalization, worktree detection, skill diagnostics, or first-party setup; the cited account-action location only prints lint findings.
- The setup prompt permits declining, non-interactive setup records no consent, and staging and analysis check the saved consent window.
- The pipeline stages transcripts locally, and the later inbox request sends the derived analysis payload rather than the raw transcript.
- No install hook or unrelated account action was established.
Source & flagged code
23 flagged · loading sourcePackage contains a critical-looking secret pattern.
tests/public_mirror_export.mjsView on unpkg · L103AWS access key ID in tests/public_mirror_export.mjs
tests/public_mirror_export.mjsView on unpkg · L103Findings
Affected versions and remediation
This report applies to @bongos/core@1.20.46.
See all versions of @bongos/core for other recorded verdicts.
- Review the evidence and your use of @bongos/core@1.20.46 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.