dsh-agy-link@0.4.70
Google Antigravity (agy CLI) models for DeepSeek Harness — stream Gemini/Claude/GPT-OSS subscriptions into DSH with thinking, tool activity, token usage and in-GUI Google OAuth login.
AI Security Review
scanned 10d ago · by lpm-firewall-aiThe default approval bypass is a dangerous capability and warrants a warning. The OAuth flow inspected targets Google and stores credentials for Antigravity, so the source does not support a malicious credential-theft finding.
Decision evidence
public snapshot- package.json routes the package entry point to dist/index.js.
- cordis.patch.yml enables the plugin by default and defaults its permission mode to skip unless DSH_AGY_MODE is set.
- dist/index.js translates skip mode into the CLI flag that bypasses tool approvals.
- The OAuth flow uses Google authorization, token, and user-info endpoints; its caller saves tokens in Antigravity's token file.
- The source exposes a local DSH-to-Antigravity integration, consistent with the package's stated provider purpose.
- The client inserts the renderer’s SVG output into a dedicated `.agy-mermaid-body` div as HTML.
- The permission-skipping default is a risky capability, but the inspected source does not establish that the package uses it for an unrelated or hidden action.
- Inspected OAuth calls send credentials to Google endpoints and persist them for Antigravity; no unrelated credential recipient was found.
Source & flagged code
7 flaggedThis package version adds a dangerous source file absent from the previous stored version; route for source-aware review.
dist/index.jsView on unpkgFindings
Affected versions and remediation
This report applies to dsh-agy-link@0.4.70.
See all versions of dsh-agy-link for other recorded verdicts.
- Review the evidence and your use of dsh-agy-link@0.4.70 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.