The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the npx mushi-mushi command to run an interactive setup wizard. This CLI tool automates framework detection, authentication via device flow, and project configuration.
[EXTERNAL_DOWNLOADS]: The setup process downloads and installs various SDK packages from the @mushi-mushi/ scope on NPM, such as @mushi-mushi/react and @mushi-mushi/node. These packages are official components of the Mushi Mushi ecosystem.
[DATA_EXFILTRATION]: The skill automates the writing of API keys and Project IDs to a local .env.local file. This is a standard and recommended practice for managing local development secrets and does not involve sending sensitive data to unauthorized third parties.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to process output from the CLI tool, specifically for copying and pasting initialization snippets into the codebase.
Ingestion points: CLI output from the setup wizard and the mushi doctor diagnostic tool described in SKILL.md.
Boundary markers: Not explicitly defined; the agent is directed to use the tool's output to complete the installation.
Capability inventory: Execution of shell commands via npx, writing to local environment files, and modifying editor configurations (MCP).
Sanitization: No specific sanitization or validation is performed on the CLI output before the agent implements the suggested code snippets.

mushi-setup