customer-workspace
Pass
Audited by Gen Agent Trust Hub on May 2, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Includes a PowerShell script (
scripts/Initialize-CustomerWorkspace.ps1) for local workspace setup. The script automates directory creation and template deployment using standard system commands. - [PROMPT_INJECTION]: The skill processes untrusted input from chat and email logs, creating a surface for indirect prompt injection where malicious instructions in the input could influence the agent's behavior during processing.
- Ingestion points:
assets/inbox.prompt.mdandassets/convert-meeting-minutes.prompt.mdingest external text content for classification and formatting. - Boundary markers: Absent; the processing prompts do not define explicit delimiters to separate agent instructions from the untrusted data.
- Capability inventory: Capability is limited to file system writes within the workspace (e.g.,
_inbox/and date-named folders). - Sanitization: No sanitization is performed on external data before it is interpolated into templates and written to workspace files.
Audit Metadata