The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the project being analyzed, including file contents and git history. This data is interpolated into prompts to generate architectural reports and used in decision-making steps without explicit sanitization or boundary markers.
Ingestion points: Project source code files, git log output, and the codebase-design configuration directory.
Boundary markers: Absent. The instructions do not specify using delimiters or instructions to ignore potential adversarial content embedded in the repository.
Capability inventory: The skill possesses capabilities for reading the filesystem, executing git log via shell, writing temporary HTML files, and triggering the system's default file opener.
Sanitization: The skill lacks validation or escaping for data harvested from the repository before it is included in the generated report or processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill generates reports that fetch assets from well-known content delivery networks (CDNs).
Evidence: The template in HTML-REPORT.md includes script tags for Tailwind CSS (cdn.tailwindcss.com) and the Mermaid visualization library (cdn.jsdelivr.net). These are trusted, well-known services used for their intended purpose of styling and rendering charts.
[COMMAND_EXECUTION]: The skill executes shell commands to gather project metadata.
Evidence: The skill uses git log --oneline to identify codebase hotspots. This is a standard development operation used here for architecture analysis.

improve-codebase-architecture