The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses bun to execute several local maintenance and production scripts, such as assets/capture.ts for rendering images and assets/verify-full-text.ts for content validation. These executions are constrained to the skill's own codebase and are essential for its operation.
[EXTERNAL_DOWNLOADS]: The skill utilizes the playwright package and its official browser installer (chromium). These are well-known tools from Microsoft (a trusted vendor) and are used strictly for the rendering process. The skill instructions also emphasize avoiding remote font loads to ensure privacy and stability.
[INDIRECT_PROMPT_INJECTION]: As the skill processes external content (URLs, pasted text) and renders it via HTML, it possesses an inherent attack surface. However, the author has implemented significant defenses: a strict block-by-block comparison tool (verify-full-text.ts) ensures the rendered output has not been modified or injected with new instructions, and the capture.ts script validates the logical structure of complex 'whiteboard' renders against a pre-defined ledger.
[OBFUSCATION]: The assets/audit.ts script uses String.fromCharCode to define search tokens (e.g., 'svg', 'npm') for its internal linting process. This is a benign implementation used to enforce technical constraints (such as preventing the use of vector assets or legacy package managers) within the skill's source files.

ljg-card