windsurf-core-workflow-b
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides workflow templates that process untrusted data (such as user-provided feature names and git diff output) and interpolate them into file paths and shell commands.
- Ingestion points: User-supplied variables in the 'new-feature' workflow; output from
git diffin the 'review-pr' workflow. - Boundary markers: The templates do not define specific delimiters or instructions to ignore embedded commands within the processed data.
- Capability inventory: The skill utilizes
Bash,Write,Edit, andReadtools for file creation, testing, and deployment automation. - Sanitization: The examples do not demonstrate validation or escaping for interpolated variables, though this is typical for documentation templates.
- [COMMAND_EXECUTION]: The skill documents the use of 'turbo' and 'turbo-all' annotations, which are used within the Windsurf environment to enable the automatic execution of shell commands such as
npm run,vitest, andgit push. - [EXTERNAL_DOWNLOADS]: The documentation references a public GitHub repository (
Windsurf-Samples/cascade-customizations-catalog) as a resource for additional workflow templates. - [DATA_EXPOSURE]: The skill references local storage paths for persistent data (
~/.codeium/windsurf/memories/) and configuration files (.windsurfrules), which is consistent with the documented feature's intended behavior.
Audit Metadata