The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/project_launcher.py utility automates environment configuration by executing pip and npm/yarn/pnpm commands via subprocess.run. It identifies backend and frontend entry points to launch services, effectively managing the lifecycle of the local development environment.
[DYNAMIC_EXECUTION]: The skill dynamically generates platform-specific launcher scripts (run_app.bat, run_app.sh, run_app.command) during project setup. It also performs automated code modifications, such as using AST analysis to rewrite Python import statements to resolve relative import issues in the target project.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to process user-provided requirements and existing codebases, the skill has an attack surface where instructions embedded in untrusted project data (like package.json scripts or requirement files) could be processed by the agent or executed via the launcher.
Ingestion points: Processes specs/requirements.md, user source code for import scanning, and configuration files such as package.json and project.config.json.
Boundary markers: Lacks explicit safety delimiters when reading or writing project files, although it follows a structured internal workflow.
Capability inventory: Possesses significant capabilities including filesystem modification, shell command execution for environment setup, and network probing for service health checks.
Sanitization: Uses AST parsing for import correction, providing a level of structural validation for code changes, but primarily relies on the integrity of the user-supplied project content.

vibe-coding-guidelines