The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted user data, such as project context, environment details, and testing requirements, to generate executable Cypress code and CI plans. It lacks explicit boundary markers or sanitization instructions to prevent instructions embedded within the user data from influencing the agent's behavior.
Ingestion points: Project context and user requirements passed to the agent via SKILL.md and prompts/ui-test-cypress.md.
Boundary markers: None identified; instructions do not use delimiters to isolate untrusted input.
Capability inventory: Generation of Javascript test files, Cypress configuration files, and shell-based CI command lines.
Sanitization: No validation or escaping of user-provided strings is specified.
[COMMAND_EXECUTION]: The skill contains scripts and instructions that trigger shell command execution. The scripts/run-tests.sh file executes npx cypress run, and evaluation fixtures in the evals/ directory utilize Python and Bash to verify outputs. While these are legitimate tools for testing, they represent a functional command execution surface.
[PRIVILEGE_ESCALATION]: The installation instructions in README.md recommend running a PowerShell script with -ExecutionPolicy Bypass. While a standard practice for local development scripts, it is a mechanism used to bypass local system security policies regarding script execution.

ui-test-cypress