The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various shell commands to manage the local development lifecycle. This includes starting a background server using bun run docs:dev, searching logs with grep and rg, querying JSON data with jq, and terminating processes using lsof and kill. These operations are scoped to the local development environment.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it automates browser interactions and processes content from a web server. While the server is local, the browser session is used to evaluate DOM states and capture logs.
Ingestion points: Web content loaded via agent-browser open from the local documentation server.
Boundary markers: None. The skill does not use specific delimiters to separate untrusted browser content from instructions.
Capability inventory: Includes browser interaction (click, fill, wait), JavaScript evaluation (eval), and local shell execution (kill, bun, jq).
Sanitization: None observed for data extracted from the browser before it is processed or used in subsequent eval calls.
[DYNAMIC_EXECUTION]: The skill uses agent-browser eval to inject and execute JavaScript snippets into the browser context. This is used for capturing runtime errors, monitoring unhandled rejections, and inspecting the semantic state of the UI (e.g., checking aria-current or data-standalone-demo attributes).

local-testing