The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates a workflow where the agent processes untrusted project data and has access to sensitive tools, creating a risk of indirect prompt injection.
Ingestion points: The skill utilizes tools such as get_file_text_by_path and search_in_files_by_text to read project source code and metadata into the agent's context (SKILL.md).
Boundary markers: There are no specific instructions or delimiters defined to help the agent distinguish between its own system instructions and potentially malicious instructions embedded within the code files it processes.
Capability inventory: The skill has extensive access to modify the filesystem via replace_text_in_file and create_new_file, and it can execute arbitrary shell commands through the execute_terminal_command tool (SKILL.md).
Sanitization: No sanitization, filtering, or validation steps are described for handling external content before it is processed or used to drive tool execution.

jetbrains-coding