The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The scripts scripts/build-graph.py and scripts/parse-js.mjs recursively read source files from the user project to build dependency graphs and symbol lists.
Boundary markers: Data is processed as raw source code; outputs do not include explicit boundary markers or warnings differentiating code structure from embedded string/comment content.
Capability inventory: The skill utilizes ast-grep (which supports file rewriting via -r) and provides test scripts that use subprocess and execFileSync to call internal helpers.
Sanitization: Content is analyzed via static parsing (Abstract Syntax Tree) rather than execution, which effectively limits the risk to output poisoning.
[EXTERNAL_DOWNLOADS]:
The SKILL.md contains instructions for installing the ast-grep utility using well-known services and package registries: npm install -g @ast-grep/cli, brew install ast-grep, and cargo install ast-grep --locked.
[COMMAND_EXECUTION]:
The skill is built to generate and execute structural search and refactoring commands using ast-grep.
Test files tests/test_build_graph.py and tests/test-parse-js.mjs utilize subprocess.run and execFileSync respectively to invoke the bundled analysis scripts during development and testing.

ast-code-graph