The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes system-level commands (open, xdg-open, or start) to display the generated HTML architecture report. This is a standard functional requirement for a reporting tool and is limited to opening a file created by the agent in a temporary directory.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it ingests and processes content from the local codebase.
Ingestion points: The agent reads project files, specifically CONTEXT.md and docs/adr/, to identify refactoring candidates.
Boundary markers: The skill enforces a strict technical vocabulary (module, interface, implementation, seam, adapter), which acts as a conceptual filter against unrelated or malicious content in the codebase.
Capability inventory: The skill is capable of reading local files, writing HTML files to system temporary directories (/tmp, $TMPDIR, %TEMP%), and invoking browser/file-opening commands.
Sanitization: The instructions explicitly mandate that the generated HTML must be self-contained and strictly prohibits the inclusion of external scripts, stylesheets, or CDNs, effectively mitigating Cross-Site Scripting (XSS) risks in the final report.

scan-architecture