The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-controlled content from PRODUCT.md, DESIGN.md, and project source files into the agent's context to guide design decisions.
Ingestion points: scripts/load-context.mjs reads product and design metadata. scripts/live-wrap.mjs and scripts/live-accept.mjs read and modify project source files.
Boundary markers: The skill uses markers like  to identify its own blocks but does not employ specific 'ignore embedded instructions' delimiters for the user-supplied strategic content in PRODUCT.md.
Capability inventory: The skill possesses powerful capabilities including subprocess execution (execSync), file system modification (writeFileSync, rmSync), and local network server creation.
Sanitization: The skill processes markdown and code strings using standard search-and-replace and regex patterns without specialized sanitization for embedded natural language instructions.
[COMMAND_EXECUTION]: The skill relies on several utility scripts to automate design tasks and environment management.
Evidence: scripts/live-poll.mjs, scripts/live.mjs, and scripts/pin.mjs utilize execSync and spawn to run Node.js helpers and npx commands. This is used for creating variant wrappers, injecting scripts into project files, and creating symlink shortcuts for the agent.
[EXTERNAL_DOWNLOADS]: The skill references well-known external services for design assets and package management.
Evidence: reference/brand.md provides instructions for fetching stock imagery via Unsplash templates. The skill's commands may also trigger npx package downloads from the official npm registry.

impeccable