The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for shell scripts (workspace-setup.sh and feature-specific E2E scripts) that use system commands like git, lsof, sed, and chmod. These scripts are intended to automate local environment configuration and service verification.
[EXTERNAL_DOWNLOADS]: The templates describe the use of curl for health checks on local endpoints and provide examples of installing project dependencies using standard package managers like npm, bun, pip, and cargo.
[PROMPT_INJECTION]: The skill defines a workflow where an agent processes data from .dev-workflow/feature-verification.json to perform verification steps, creating an indirect injection surface.
Ingestion point: .dev-workflow/feature-verification.json (referenced in the Evaluator Integration section).
Boundary markers: No specific delimiters or "ignore instructions" warnings are defined for the agent when it interprets verification steps from these files.
Capability inventory: The agent is instructed to execute shell scripts, use curl, and utilize agent-browser based on the logic defined in the processed files.
Sanitization: No explicit validation or sanitization of the JSON content is mentioned in the setup instructions.

aep-testing-guide