The Agent Skills Directory

[SAFE]: The skill establishes a comprehensive set of markdown-based guidelines for web development tasks. It does not contain any executable scripts, obfuscated payloads, or commands that interact with the host system in an unsafe manner. The design emphasizes reusing existing project components and verifying outputs in their real product context.
[SAFE]: The skill references external repositories github.com/kostja94/bricks and github.com/kostja94/contextus for UI components and project context management. These are documented vendor resources belonging to the skill author and are used legitimately for construction guidance.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading project-specific files (e.g., contextus.md, pages.json) and external documentation to generate code and content, which creates a potential surface for indirect prompt injection from untrusted data sources.
Ingestion points: The references/workflow.md (Step 1) and references/contextus.md define a process for reading contextus.md and other project context modules to determine page requirements and content.
Boundary markers: The references/workflow.md includes a Step 2 to "Define The Page Contract" and Step 7 to "Verify In Context," which act as manual and automated validation steps. The references/quality-gates.md provides a checklist for manual review.
Capability inventory: The skill allows the agent to create new pages, route handlers, and data adapters within the target repository.
Sanitization: The skill instructs the agent to use "authoritative fields" and "real product evidence," and explicitly warns against inventing product claims or using generic filler in references/quality-gates.md and SKILL.md boundaries.

page-builder