The Agent Skills Directory

[PROMPT_INJECTION]: The skill dynamically reads and executes instructions from markdown files, creating a surface for indirect prompt injection.\n
Ingestion points: The skill searches for .md files in project-local ./agents/ and global ~/.claude/agents/ directories (or user-defined paths).\n
Boundary markers: Persona instructions from the discovered files are concatenated directly into the sub-agent prompt without the use of delimiters or protective instructions (e.g., "the following content is data, not instructions").\n
Capability inventory: The skill uses the Agent tool to execute the resulting prompts, granting the sub-agents any capabilities associated with that tool.\n
Sanitization: No validation or filtering is performed on the file content before it is processed.

team-builder