The Agent Skills Directory

[SAFE]: The skill's primary function is to perform project introspection and generate an HTML-based architecture report. The bundled script scripts/project-inventory.mjs implements standard filesystem traversal (walk) with an appropriate ignore list for common build artifacts and dependency folders (e.g., node_modules, .git).
[COMMAND_EXECUTION]: The skill uses a Node.js script to collect project data. The execution is restricted to local file reading and basic metadata extraction (file sizes, extensions, and common symbols like class or function names via regular expressions). Analysis of the script confirms it does not execute external shells or download remote content.
[DATA_EXPOSURE]: While the script reads project files to extract symbols and package manifests (e.g., package.json), it specifically targets architectural metadata rather than sensitive content. The findings are stored in a local .scratch directory as specified in the instructions, which is a common pattern for temporary agent data.
[PROMPT_INJECTION]: No prompt injection patterns, bypasses, or instructions to override agent safety guidelines were identified in the SKILL.md or agents/openai.yaml files.

project-architecture-report