The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using curl to interact with a local notification service (http://localhost:8888/notify). This is used solely for voice notifications to update the user on the progress of CLI generation and modification workflows.\n- [COMMAND_EXECUTION]: The skill performs validation of generated code by setting executable permissions using chmod +x and executing the resulting script with the --help flag. This ensures the output is functional before reporting success to the user.\n- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it is designed to load and apply instructions from a local customization directory (~/.claude/PAI/USER/SKILLCUSTOMIZATIONS/CreateCLI/).\n
Ingestion points: The SKILL.md file specifies that the agent should load and apply PREFERENCES.md or other configuration files from a user-controlled local path.\n
Boundary markers: Absent. The skill does not implement specific delimiters or 'ignore' instructions to isolate the content of customization files from the agent's core operational logic.\n
Capability inventory: Across its workflows, the skill can execute shell commands (curl, chmod, script execution), perform file system operations (read/write/ls), and generate new executable scripts.\n
Sanitization: Absent. Information retrieved from the customization files is integrated into the agent's context and logic without explicit validation, filtering, or sanitization.

CreateCLI