The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent clones and processes external mobile app repositories (referenced as vars.APP_REPO). This establishes a data ingestion surface where malicious instructions or code embedded within the target repository could influence the agent's behavior during the build, codegen, or analysis phases.
Ingestion points: External repositories cloned via GitHub Actions in Section 8.
Boundary markers: None specified for the processed repository content.
Capability inventory: Shell command execution (run blocks), file system access (cloning to ../app), and network operations (Playwright, git).
Sanitization: No explicit sanitization or validation of the external repository's content is described.
[COMMAND_EXECUTION]: The skill provides several shell command templates for performance measurement and CI/CD automation. These include using curl to inspect headers and pnpm to install dependencies.
Evidence: Section 1 contains curl commands for gzipped content inspection. Section 8 contains GitHub Actions YAML for installing Playwright and cloning repositories.
[DYNAMIC_EXECUTION]: The skill uses Playwright's page.evaluate to execute arbitrary JavaScript within the context of the browser during automated screen capture and testing.
Evidence: Section 4 uses page.evaluate to dispatch mouse events to the accessibility tree.

mobile-app-to-web-demo