The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions frequently use local script execution commands (e.g., node scripts/browser.js) to drive browser automation and interact with the dashboard API.
[INDIRECT_PROMPT_INJECTION]: The skill directs the agent to read and follow guidance embedded in the webpage DOM via the [data-testid="llm-instructions"] element. This creates a surface where external content could influence agent behavior.
Ingestion points: The agent reads content from document.querySelector('[data-testid="llm-instructions"]').textContent as described in SKILL.md.
Boundary markers: None identified; the agent is instructed to read and use the text directly.
Capability inventory: The skill uses playwright-cli and node to execute eval statements and shell commands.
Sanitization: None identified; the instructions are treated as valid guidance.
[DYNAMIC_EXECUTION]: The skill's core functionality relies on generating and executing JavaScript snippets via eval within the context of a browser session to perform data operations.

agent-desk