The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates an architectural review report as an HTML file in the system's temporary directory ($TMPDIR) and uses platform-specific commands (xdg-open, open, or start) to open it for the user.
[EXTERNAL_DOWNLOADS]: The generated HTML report includes scripts to load Tailwind CSS from cdn.tailwindcss.com and Mermaid.js from cdn.jsdelivr.net. These are well-known, trusted CDNs used for rendering the report's styling and diagrams.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted codebase content to perform its analysis.
Ingestion points: Codebase files, module lists, and commit history are ingested via /code-graph queries, git log, and grep commands across multiple files including SKILL.md and sweep-patterns.md.
Boundary markers: Absent; there are no explicit delimiters or instructions to ignore embedded commands within the analyzed code snippets.
Capability inventory: The skill can execute shell commands (git, grep, xdg-open), write to the local filesystem (HTML report, CONTEXT.md), and spawn sub-agents.
Sanitization: No explicit sanitization or escaping of the codebase content is performed before it is interpolated into agent prompts or the HTML report.
[DATA_EXPOSURE]: The 'Security agent' defined in sweep-patterns.md is explicitly tasked with identifying hardcoded secrets and unguarded environment variable reads within the codebase. This behavior is transparently documented and aligns with the skill's stated purpose of codebase health and architectural review.

improve-codebase-architecture