The Agent Skills Directory

[COMMAND_EXECUTION]: The skill defines several CLI-based operations for the agent to execute, such as querying symbol definitions, call chains, and impact analysis using the codegraph utility.
[EXTERNAL_DOWNLOADS]: The command resolution logic includes fallback mechanisms using bunx and npx to fetch and execute the codegraph and @terrain-ai/cli packages from the NPM registry if they are not available in the local directory.
[INDIRECT_PROMPT_INJECTION]: The skill processes project source code to generate and query an AST graph. While it does not directly ingest untrusted network data, the reading of local codebase content presents a surface where malicious source code or comments could attempt to influence agent responses during code analysis.

codegraph-skill