The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation includes instructions to install a plugin from an external source: /plugin install team-agents@duyet-claude-plugins. This resource is associated with the author's namespace and is part of the intended setup for team coordination functions.\n- [COMMAND_EXECUTION]: The skill architecture spawns worker agents that are explicitly authorized to use system tools, including Bash and file-system commands (Read, Write, Edit), which are necessary for the primary purpose of task execution and codebase manipulation.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it ingests untrusted codebase data and user requests to generate instructions for sub-agents.\n
Ingestion points: Data is ingested when worker agents read local source code, analyze pull requests, or process external documentation during research phases (e.g., in SKILL.md and references/domains/research.md).\n
Boundary markers: The skill implements a 'WORKER preamble' designed to constrain sub-agents and define their operational boundaries.\n
Capability inventory: Delegated worker agents have broad system access, including Bash execution and file-system write capabilities, to implement features or fix bugs across the project files.\n
Sanitization: The logic relies on descriptive prompt instructions to maintain safety and does not include programmatic sanitization or filtering of the content ingested from files or external requests before interpolation into sub-agent prompts.

orchestration