The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to execute shell commands (xdg-open on Linux, open on macOS, or start on Windows) to open the generated HTML report. While the file path includes a user-provided name, the instructions specifically require sanitizing the name for the filesystem and quoting the absolute path to mitigate command injection risks.
[EXTERNAL_DOWNLOADS]: When generating the optional HTML report, the skill fetches Tailwind CSS from https://cdn.tailwindcss.com and Mermaid.js from https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.esm.min.mjs. These are well-known technology services used for layout and visualization.
[PROMPT_INJECTION]: The skill processes untrusted user-provided life context from CONTEXT.md. This represents an indirect prompt injection surface as this data is later interpolated into the agent's interpretation and the HTML report.
Ingestion points: persons/<name>/CONTEXT.md (read in references/artifacts.md).
Boundary markers: Not explicitly specified for the content of the context file.
Capability inventory: The skill can write files and execute shell commands (xdg-open, etc.) to display results.
Sanitization: Instructions in references/html-report.md mandate escaping artifact values before HTML insertion and sanitizing filenames.

parashari-judgement