The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to perform administrative and development tasks including building software from source and managing terminal multiplexers. It specifically provides recipes for using tmux to run commands, send keyboard input, and capture pane output for analysis.
[REMOTE_CODE_EXECUTION]: For web-based application testing, the skill directs the agent to execute JavaScript via eval within a browser context (CDP) to interact with the application's internal state and simulated UI actions.
[CREDENTIALS_UNSAFE]: The documentation mentions the requirement for credentials and tokens in the testing 'Pre-state' to facilitate authenticated access during scenarios. It advises on using symlinks to manage these inputs while keeping them separate from mutable state.
[PROMPT_INJECTION]: The skill is designed to ingest and act upon 'scenario cards' (external Markdown files). Because these files contain natural language instructions that the agent is expected to follow, they represent an indirect prompt injection surface.
Ingestion points: External .md scenario card files referenced in the workflow.
Boundary markers: The skill does not define specific delimiters or warnings to prevent the agent from obeying malicious instructions embedded within a scenario card.
Capability inventory: The agent has the ability to execute shell commands, compile binaries, and execute JavaScript in a browser.
Sanitization: There are no instructions for validating or sanitizing the content of the scenario cards before processing.
[NO_CODE]: This skill consists solely of instructional documentation in Markdown format and does not include any standalone scripts, binaries, or automated configuration files.

e2e-scenario-testing