The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands using user-provided input without evident sanitization or validation, creating a risk for command injection.
Evidence: The workflow in SKILL.md executes npx playwright test --project="chromium:regression" {feature}.spec.ts where {feature} is derived directly from the user's request.
[DYNAMIC_EXECUTION]: The skill follows a pattern of generating executable code and then immediately running it in the local environment.
Evidence: The SKILL.md file describes a workflow to create spec files (e.g., e2e/tests/regression/{feature}.spec.ts) using templates and then executing them via the playwright test runner.
[INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface for indirect prompt injection as it interpolates untrusted user input into both files and commands.
Ingestion points: User-provided feature-name from the request or argument-hint in SKILL.md.
Boundary markers: None present; the skill lacks delimiters or instructions for the agent to ignore malicious patterns in the input.
Capability inventory: The skill has file-write capabilities for POMs, test data, fixtures, and spec files, as well as shell command execution capabilities via npx.
Sanitization: There is no logic described to sanitize the feature-name before using it in file paths or shell commands.

create-regression-test