The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a multi-agent architecture where a 'Lead' agent monitors and acts upon data generated by 'Teammate' agents.
Ingestion points: As described in references/agent-teams-guide.md, data enters the agent context through a shared 'Task list' and a 'Mailbox' system for inter-agent messaging.
Boundary markers: The instructions do not provide explicit delimiters or warnings to the Lead agent to treat teammate outputs as untrusted data, increasing the risk of the Lead following instructions embedded in teammate reports.
Capability inventory: Teammates are described as full Claude Code instances with access to the project context and relevant tools.
Sanitization: The skill does not provide mechanisms for sanitizing or validating messages exchanged between teammates and the Lead.
[EXTERNAL_DOWNLOADS]: The documentation references and encourages the installation of a third-party tool from an unverified external source to support specific functionality.
Evidence: references/agent-teams-guide.md instructs users to install the it2 CLI from a GitHub repository (https://github.com/mkusaka/it2) to enable split-pane mode in iTerm2.

claude-code-agent-teams