The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx extract-design-system to download and execute a package from the npm registry. As this package is not from a verified or well-known vendor, it represents a risk of arbitrary code execution on the host machine.
[EXTERNAL_DOWNLOADS]: The workflow requires installing the Chromium browser via the Playwright framework (npx playwright install chromium) and fetching the extract-design-system package, which involves downloading external code and binaries at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes and reviews data extracted from user-provided website URLs, creating a surface for indirect prompt injection where malicious content on a website could attempt to influence the agent's behavior.
Ingestion points: The agent ingests data from public websites via the extraction tool (SKILL.md).
Boundary markers: The instructions include warnings to treat extracted output as untrusted third-party input and to seek user confirmation before modifying project files (references/workflow.md).
Capability inventory: The agent has the ability to execute shell commands via npx and write JSON and CSS files to the local directory.
Sanitization: There is no explicit sanitization or filtering of the website content before it is parsed and presented to the agent for analysis.

extract-design-system