The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user drafts and interpolates them directly into an HTML template without sanitization, creating a surface for HTML injection. * Ingestion points: User-provided draft text (via paste or file path) as defined in the 'Inputs' section of SKILL.md. * Boundary markers: Absent; the content is placed directly into the placeholder in the HTML template. * Capability inventory: Uses the 'Write' tool to create a local file and the 'Bash' tool to execute 'open', launching the file in a web browser. * Sanitization: Absent; there are no instructions to escape HTML tags or sanitize user input before inclusion in the generated view.
[COMMAND_EXECUTION]: The skill uses the system 'open' command to launch the generated HTML file in the default browser. * Evidence: The skill workflow in 'SKILL.md' explicitly calls 'open _private/views/{name}-comments-temp.html'.

comment-mode