@sous-io/sous@1.0.2
Compiles AI coding agent configuration (CLAUDE.md, skills, memories) from LiquidJS templates
AI Security Review
scanned 5d ago · by lpm-firewall-aiThe bundle has a credential forwarding flaw: it follows cross-origin redirects while retaining the repository token header. This is a serious vulnerability in repository fetching, though the source does not show a fixed malicious collector or deliberate credential theft.
Decision evidence
public snapshot- The HTTP read strategy adds a repository token as a Bearer Authorization header.
- Each redirect request reuses the same headers, including the Authorization header.
- Redirect destinations are taken from the response Location without checking that the origin stays the same.
- GitHub repository reads can obtain tokens from environment variables, Git credentials, or the GitHub CLI.
- The package describes itself as a CLI for compiling AI agent configuration and has no install or postinstall hook.
- Fetching configured repositories with their credentials is consistent with the package’s purpose; the issue is that credentials can carry across redirects.
Source & flagged code
11 flaggedA single source file combines environment access, network access, and code or shell execution with blocking evidence.
dist/sous.jsView on unpkg · L731Findings
Affected versions and remediation
This report applies to @sous-io/sous@1.0.2.
See all versions of @sous-io/sous for other recorded verdicts.
- Review the evidence and your use of @sous-io/sous@1.0.2 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.