The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/aidr executes the aider CLI tool using subprocess.run with a list of arguments, which is a secure method that prevents shell injection attacks.
[EXTERNAL_DOWNLOADS]: The scripts/setup.sh script installs the aider-chat package from PyPI via the uv tool. This is a standard installation procedure for a well-known and reputable developer tool.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted data from the repository into its operations. \n * Ingestion points: The skill reads the contents of the target repository's files and accepts a user-provided goal string, both of which are passed to the underlying AI model. \n * Boundary markers: No delimiters or specific safety instructions (e.g., "ignore embedded commands") are applied to the user-provided goals before they are processed by Aider. \n * Capability inventory: The Aider tool has the capability to read and modify files in the local filesystem, presenting a risk if an injection occurs. \n * Sanitization: The skill does not perform validation or sanitization of the input goals or the codebase content.

aidr