fast-context
Fail
Audited by Gen Agent Trust Hub on Jun 5, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill contains logic in
scripts/lib/extract-key.mjsto locate and read the Windsurf application's local SQLite database (state.vscdb). It extracts sensitive authentication tokens (apiKey) and provides CLI flags such as--print-keythat expose these credentials in plaintext. - [COMMAND_EXECUTION]: The
ToolExecutorclass inscripts/lib/executor.mjsexecutes local file system operations based on commands received from a remote API. The path resolution implementation in the_realmethod fails to sanitize against path traversal (e.g.,..sequences), allowing the remote service to access files outside the intended project directory. - [DATA_EXFILTRATION]: The skill's communication logic in
scripts/lib/core.mjsautomatically disables TLS certificate verification (NODE_TLS_REJECT_UNAUTHORIZED = '0') if an initial connection attempt fails. This exposes sensitive codebase data and authentication tokens transmitted to the remote API to man-in-the-middle attacks. - [SAFE]: The skill utilizes the Microsoft-maintained
@vscode/ripgreppackage for optimized local searching. - [SAFE]: Network operations are directed to
windsurf.com(operated by Codeium), which is a recognized service in the AI development tool ecosystem.
Recommendations
- AI detected serious security threats
Audit Metadata