@ngockhoale/ukit@3.4.60
Install/update an index-first AI workspace for Claude Code, OpenAI Codex and omp (Oh My Pi).
AI Security Review
scanned 20h ago · by lpm-firewall-aiThe source confirms that Claude credentials can be forwarded to a separate, package-default decision endpoint without a decision-specific key. The provider is presented as part of UKit's product, so the source supports a warning for credential exposure rather than a confirmed malicious transfer.
Decision evidence
public snapshot- The decision client defaults to a separate UNIC endpoint, while also allowing an environment or saved gateway URL to take precedence.
- If no decision-specific key is set, the resolver falls back to the shared gateway credential; the shared resolver accepts Claude API keys and auth tokens.
- The client places the resolved key in an authorization header and sends it in a POST request to the resolved endpoint.
- Runtime decision calls require the decision stage to be configured as non-off. UKit describes this as a separate product decision lane, but source does not establish recipient-specific consent for forwarding a Claude credential.
- The credential transfer is conditional on a decision request and an available shared credential; the source does not establish how the recipient uses it.
Source & flagged code
11 flagged · loading sourcePackage source references dynamic require/import behavior.
src/core/runtimeConfig.jsView on unpkg · L9Source file is highly similar to a previously finalized malicious package; route for source-aware review.
scripts/release/verify-release.mjsView on unpkgFindings
Affected versions and remediation
This report applies to @ngockhoale/ukit@3.4.60.
See all versions of @ngockhoale/ukit for other recorded verdicts.
- Review the evidence and your use of @ngockhoale/ukit@3.4.60 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.