The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides documentation with shell command examples for file management (mkdir, ln, echo) and environment configuration (export). These are standard workflows for tool configuration and project management and present no inherent risk.\n- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a process where the agent reads instructions from skill files and updates them based on session history. This creates a surface for context-based instruction processing. \n
Ingestion points: Skill files in the .claude/skills/ directory and conversation history.\n
Boundary markers: Not explicitly defined; relies on platform-level agent context management.\n
Capability inventory: The skill uses Bash, Read, Grep, and Glob to modify and manage local files.\n
Sanitization: The skill mitigates risks by recommending Git-based version control and explicit review sessions before finalizing updates.\n- [EXTERNAL_DOWNLOADS]: The skill references external Git repositories on well-known platforms like GitHub for collaborative skill sharing. These references are illustrative of standard team workflows and target trusted services.

claude-collaboration