The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill dynamically generates a JavaScript file (screenshot-script.mjs) and executes it using the node runtime. The content of this script, specifically the SCREENSHOTS configuration and AUTH object, is constructed at runtime using data gathered from the user and the local codebase environment.
[CREDENTIALS_UNSAFE]: The skill explicitly instructs the agent to solicit sensitive authentication details (email/username and password) from the user. These credentials are subsequently embedded in plain text within the generated screenshot-script.mjs file during the execution phase.
[COMMAND_EXECUTION]: The skill performs several shell-based operations including environment validation (npx playwright), directory creation (mkdir -p screenshots), execution of the generated script (node screenshot-script.mjs), and file cleanup (rm screenshot-script.mjs).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated analysis of untrusted codebase files.
Ingestion points: Reads content from README.md, CHANGELOG.md, config/routes.rb, app/ directories, and other routing/component files to identify app features.
Boundary markers: None identified; instructions do not specify delimiters or warnings to ignore embedded content in analyzed files.
Capability inventory: The skill has the capability to write and execute scripts (node), create directories, and perform network requests via Playwright.
Sanitization: There is no evidence of sanitization or escaping of the paths or feature names extracted from the codebase before they are interpolated into the generated JavaScript template.

screenshots