wspace
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill features a "Goal loop" that autonomously fetches and executes tasks from external GitHub issues via
gh search issues. This pattern relies on external data to influence the agent's workflow. Evidence: (1) Ingestion points: Issue titles and repository metadata retrieved inSKILL.md(Goal loop). (2) Boundary markers: The skill defines workflow stopping points (e.g., human-in-the-loop for deployments) but does not provide content delimiters or instructions to ignore embedded prompts in ingested data. (3) Capability inventory: Full shell access forgit,gh,npm, anddenoacross multiple repositories. (4) Sanitization: No explicit validation or sanitization of issue content is mentioned. - [DATA_EXFILTRATION]: The skill handles sensitive information by synchronizing environment files from a central
secrets/directory into repository checkouts using thewspace env synccommand. - [COMMAND_EXECUTION]: The skill automates the execution of local repository commands, including
npm run precommitanddeno task ci, to validate changes during the development pipeline.
Audit Metadata