The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Dynamically downloads and executes the Google Lighthouse package from the npm registry using npx. As Lighthouse is a well-known service from a trusted vendor (Google), this behavior is considered safe but is documented as an external dependency.
[COMMAND_EXECUTION]: The skill instructions include shell commands for running audits (npx lighthouse <url>) and viewing reports. While necessary for the skill's function, this relies on the underlying agent to safely handle the URL parameter and execution environment.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) because it ingests untrusted data from the filesystem and the web while maintaining high-privilege capabilities.
Ingestion points: Reads local code via Read and Glob tools; fetches remote content via WebFetch and WebSearch tools.
Boundary markers: None present. There are no instructions to the agent to disregard or isolate embedded commands in the data it analyzes.
Capability inventory: High-risk tools are available, including Bash for command execution and Write/Edit for file system modification.
Sanitization: No explicit sanitization or validation of the input data is specified before it is processed by the agent.

agent-ux-optimizer