The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from user-provided target URLs during the testability assessment.
Ingestion points: The agent navigates to external web addresses provided in the TEST_URL environment variable or as script arguments.
Boundary markers: No specific boundary markers or instructions to ignore embedded content within the analyzed pages are provided in the skill text.
Capability inventory: The skill utilizes Playwright and Vibium for browser automation, generates HTML/JSON reports, and executes shell scripts.
Sanitization: No explicit sanitization or validation of the web content is described before it is scored and incorporated into reports.
[EXTERNAL_DOWNLOADS]: The skill instructs the user/agent to install and execute external code using package managers.
Evidence: The documentation suggests running claude mcp add vibium -- npx -y vibium, which downloads and executes the vibium package from the npm registry at runtime.
Evidence: Uses npx playwright test which may trigger downloads of browser binaries or package updates.
[COMMAND_EXECUTION]: The skill uses shell script wrappers to perform assessments, which can be vulnerable to argument injection.
Evidence: The command .claude/skills/testability-scoring/scripts/run-assessment.sh https://example.com/ passes a URL directly to a shell script. If the URL is not properly quoted or sanitized within the script, it could lead to command injection.
[DYNAMIC_EXECUTION]: The skill generates and opens reports automatically, involving runtime file system operations and potentially browser-based execution of generated code.
Evidence: The assessment workflow includes generating HTML reports with radar charts and automatically opening them in a browser (AUTO_OPEN=true).

qe-testability-scoring