The Agent Skills Directory

[SAFE]: The skill consists of documentation and usage examples for the ast-grep command-line tool. It is designed to help the agent explore codebases by providing structural outlines (imports, exports, functions, etc.).
[COMMAND_EXECUTION]: The instructions involve executing the ast-grep utility locally on the file system. These commands are used for reading and mapping code structure, not for modifying the system or executing arbitrary code. One example uses git diff --name-only to identify files for analysis, which is a standard development workflow.
[DATA_EXPOSURE]: The tool accesses local source code files to generate outlines. It does not access sensitive directories (like .ssh or .aws), environment variables, or hardcoded credentials. There are no network operations to send data to external servers.
[INDIRECT_PROMPT_INJECTION]: The skill processes local source code provided in the project environment. While these files are external inputs, the ast-grep outline tool performs syntax-only analysis and the skill does not interpolate the content into sensitive agent instructions or system prompts. Boundary markers are absent, but the tool's focused purpose limits the risk. (Severity: LOW)

ast-grep-outline