The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill enables the modification of files on the local filesystem using strings provided to the run function. If the agent is instructed to perform edits based on untrusted external data, this could lead to unauthorized file modifications or code injection.
Ingestion points: The path, old_str, and new_str arguments in the run function within src/edit/__init__.py.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the skill body or logic.
Capability inventory: The skill has full read and write access to files accessible by the agent, including the ability to expand user directories (e.g., ~/.ssh) via pathlib.Path.expanduser() in src/edit/__init__.py.
Sanitization: The implementation does not include sanitization, escaping, or validation of the new_str content before writing it to the target file.

edit