The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill uses npx to download and execute the @playwright/cli package from the NPM registry during the analysis workflow. Playwright is a well-known service maintained by Microsoft, a trusted organization.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using playwright-cli to navigate to web routes, resize viewports, and capture DOM snapshots and console logs.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external content including local source code (route files, components, and schema) and live web application data (DOM and console output). The ingestion of these untrusted external inputs creates a potential surface for indirect prompt injection.
Ingestion points: Reads project files such as routes, component code, and database schemas (SKILL.md Step 1); captures screenshots and console output from a target web application (SKILL.md Step 2a).
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within the analyzed data.
Capability inventory: The agent is empowered to execute shell commands (via Playwright) and write to the local filesystem (screenshots and code diffs).
Sanitization: There is no evidence of filtering or sanitization of the content retrieved from external sources before it is processed by the agent.

enhance-web-ux