The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands to manage development servers and perform browser-based verification.
Evidence: Executes startup commands such as npm run dev, yarn dev, pnpm dev, and bun dev to ensure the target application is reachable.
Evidence: Uses curl to perform health checks on the local service at http://localhost:3000.
Evidence: Employs playwright-cli to perform automated browser actions including open, snapshot, screenshot, click, fill, and console logging.
[PROMPT_INJECTION]: The skill processes a 'Sprint Contract' that defines the test plan and interaction parameters, which represents a surface for indirect instructions.
Ingestion points: Data is ingested from the 'Sprint Contract' block described in SKILL.md to dictate navigation and form interactions.
Boundary markers: Absent; the skill does not use specific delimiters to isolate the ingested contract instructions from the agent's core protocol.
Capability inventory: High-privilege browser automation (clicks, form fills, navigation) and shell command execution.
Sanitization: Input values from the contract are used directly for automation without explicit validation or escaping within the skill's logic.

evaluator