The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by ingesting untrusted data from target websites that is reflected in generated HTML reports without sanitization.\n
Ingestion points: Target website URLs and DOM metrics are extracted in testability-scoring.spec.template.js and via Vibium commands in SKILL.md.\n
Boundary markers: Absent; the skill does not implement delimiters or instructions to ignore embedded instructions in the ingested data.\n
Capability inventory: Use of child_process.exec in scripts/generate-html-report.js and testability-scoring.spec.template.js for automation.\n
Sanitization: The HTML report generator in scripts/generate-html-report.js directly interpolates metadata such as the target URL and principle descriptions into a template, creating a Cross-Site Scripting (XSS) risk if a malicious page is assessed.\n- [COMMAND_EXECUTION]: The skill utilizes shell command execution to automate the assessment and reporting lifecycle.\n
Evidence: scripts/generate-html-report.js uses child_process.exec to launch the system's default browser to display the report.\n
Evidence: resources/templates/testability-scoring.spec.template.js executes the report generation script through child_process.exec upon completion of tests.\n- [DATA_EXFILTRATION]: Local network exposure of assessment reports.\n
Evidence: scripts/generate-html-report.js creates an HTTP server that binds to 0.0.0.0. This configuration exposes the report to all devices on the local network, which may lead to unintended information disclosure.

testability-scoring