The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill templates for Blade views use the {!! !!} syntax to render the message field for dashboard tabs. This directive disables Blade's automatic HTML escaping. If a user provides malicious HTML or instructions to the agent that are then interpolated into this field, it could result in a Cross-Site Scripting (XSS) vulnerability in the generated application. While the skill documentation includes a warning to "only use trusted HTML," it does not provide or enforce a sanitization mechanism.
Ingestion points: User-supplied content intended for the __MESSAGE_HTML__ or __TAB_MESSAGE_HTML__ placeholders in SKILL.md templates.
Boundary markers: Absent. There are no delimiters or specific instructions to the agent to sanitize or escape user-provided content before interpolation.
Capability inventory: The skill generates PHP and Blade code that renders content directly to the application UI.
Sanitization: Absent. The skill relies on the agent following a text-based warning rather than implementation of a programmatic filter.
[DATA_EXPOSURE]: The skill explicitly references absolute local file paths (/home/mwguerra/projects/mwguerra/claude-code-plugins/filament-specialist/skills/docs/references/...) for documentation. While these paths are used for context, they reveal the internal directory structure of the author's workstation.

dashboard