The Agent Skills Directory

[COMMAND_EXECUTION]: Uses the shell to create project directories and initialize git repositories (mkdir, cd, git init) based on user-defined paths and settings.
[DATA_EXFILTRATION]: Manages project state and user preferences using local files such as .claude/studio-startup.local.md and .studio-startup-state.json. No unauthorized external network communication or sensitive data exposure was detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied project visions and requirements documents, which are then passed as context to downstream code-generation sub-agents.
Ingestion points: User-provided project ideas and specification files located in the docs/ideation/ directory.
Boundary markers: Requirements are interpolated into sub-agent prompts using placeholders (e.g., [requirements summary]).
Capability inventory: The skill can execute file system commands and invoke external agents with code-writing capabilities.
Sanitization: The skill relies on standard prompt engineering for context passing without explicit sanitization routines for the ingested data.

studio-startup