The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from a local software project, which constitutes untrusted data, and it possesses significant execution capabilities.
Ingestion points: Untrusted data enters the agent context through tools such as get_file_text_by_path, search_in_files_by_text, search_in_files_by_regex, and get_file_problems (found in references/jetbrains-skill.md).
Boundary markers: The skill instructions do not specify the use of delimiters or specific warnings to ignore embedded instructions within the processed project files.
Capability inventory: The skill provides access to high-privilege tools including execute_terminal_command, execute_run_configuration, create_new_file, and replace_text_in_file (found in references/jetbrains-skill.md).
Sanitization: There is no mention of sanitization, filtering, or escaping of external content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill exposes tools that allow the agent to execute shell commands and run configurations within the integrated development environment.
Evidence: The tools execute_terminal_command and execute_run_configuration are documented in references/jetbrains-skill.md and referenced in the workflow decision tree in SKILL.md.
Context: These capabilities are core to the skill's purpose as an IDE integration. The skill documentation includes warnings that these are "high-risk" and "potentially destructive," instructing the agent to require explicit user confirmation unless "Brave Mode" (an IDE-side setting) is specifically enabled.

jetbrains-skill