box-team-management
Pass
Audited by Gen Agent Trust Hub on Apr 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The instructions direct the agent to utilize the Box CLI for platform interactions if the tool is available. This is a standard and documented method for the skill's file-management purpose.
- [PROMPT_INJECTION]: The skill ingests data from user input and Markdown files stored on Box. This creates an indirect prompt injection surface typical for agents that process external content, but the logic is constrained to task tracking and does not involve risky automated operations.
- Ingestion points: Data is retrieved from
TASKS.mdandTEAM.mdon Box and provided via user interaction during setup. - Boundary markers: None. There are no specific delimiters defined to separate ingested data from agent instructions.
- Capability inventory: File reading and writing on Box (via API or CLI) and local temporary file management in a
.tmp/directory. - Sanitization: None. The agent relies on the structure of the Markdown files without explicit validation of the content.
Audit Metadata