@bongos/core@1.20.8
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-aiThe package is not shown to disclose data during installation, and its analysis path is gated by a recorded consent choice. I rate it suspicious because the analyzer sends transcript turns to the Anthropic CLI provider while setup says the raw transcripts stay local.
Decision evidence
public snapshot- The full manifest has no install lifecycle scripts or runtime self-dependency; it declares a server entry point and a CLI.
- Setup asks whether to analyze onboarding logs, but treats an empty answer as yes and says raw transcripts stay on the machine. The analyzer has a consent-window gate, with a force override.
- The analyzer builds a model prompt from transcript turns. Its default model provider is the Anthropic CLI, and the grader writes the prompt to that subprocess.
- The flagged skill-lint locations read a touched skill file and print lint results; they do not perform an account action.
- No install-time trigger or hidden install-time disclosure is established. The transcript analysis is conditional on setup consent and running the analyzer.
- The analyzer posts a derived friction report to the GDS inbox; the identified disclosure risk is the transcript prompt sent to the model provider.
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@1.20.8.
See all versions of @bongos/core for other recorded verdicts.
- Review the evidence and your use of @bongos/core@1.20.8 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.