team-setup
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs shell operations including git clone, git init, and directory creation (mkdir) to set up the team repository. Evidence found in references/mode-clone.md and references/mode-scaffold.md.
- [INDIRECT_PROMPT_INJECTION]: The skill modifies the project's AGENTS.md file to inject instructions that guide agent behavior in subsequent sessions. It uses explicit boundary markers () to delimit these instructions, as seen in references/post-setup-agents.md and the helper scripts.
- [DYNAMIC_EXECUTION]: The helper scripts (scripts/bash/team-helpers.sh and scripts/powershell/team-helpers.ps1) use Python execution via heredocs and command-line arguments to handle JSON parsing and structured file updates. This logic is used for local configuration management rather than executing remote code.
- [SAFE]: The skill implements proactive security measures, including strict validation of user-supplied paths and URLs to prevent injection attacks (e.g., blacklisting shell metacharacters), and enforces best practices for secret management by ensuring .mcp.json uses environment variable indirection.
Audit Metadata