The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves reading and interacting with the UI of a web application in 'Preview mode' using browser automation tools like read_page and find.
Ingestion points: Untrusted data from the application's DOM, console logs, and network response bodies are loaded into the agent's context during test execution in test-execution.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when processing content retrieved from the browser.
Capability inventory: The agent can perform browser automation (clicks, keyboard input, navigation), write files to the project directory (results/ folder), and execute shell commands for token processing and workspace management.
Sanitization: No filtering or sanitization procedures are defined for content read from the web application before it is evaluated by the model.
[COMMAND_EXECUTION]: The skill utilizes shell commands for administrative tasks and data processing.
The preview-access.md file provides a specific bash command for decoding JWT segments using tr and base64 for token validation.
The SKILL.md file references a migration script, scripts/migrate-workspace.py, for workspace management tasks.
The test-wizard.md file involves running git commands (e.g., git rev-parse) to establish a coverage baseline and sync point for test maintenance.

testing