The Agent Skills Directory

[COMMAND_EXECUTION]: The core logic in scripts/preflight.mjs uses spawnSync with shell: true to execute arbitrary strings defined as fixCmd or check.cmd in the profile JSON files. This allows the execution of arbitrary shell commands like pnpm, winget, and node.
[DYNAMIC_EXECUTION]: The skill executes dynamic Node.js code snippets via node -e to perform system checks. These snippets are stored in the profiles/ configuration and executed at runtime.
[PRIVILEGE_ESCALATION]: The skill facilitates the installation of system-level software, such as Visual Studio Build Tools, using winget. These operations typically require administrative privileges and modify the underlying operating system environment.
[INDIRECT_PROMPT_INJECTION]: The commandSucceeds check for better-sqlite3-native in profiles/deepworks.json attempts to import and execute a JavaScript module located within the repository being scanned (./scripts/lib/electron-native-dependencies.mjs). A malicious repository could provide a compromised version of this file to achieve code execution when the user runs the bootstrap tool.
Ingestion points: The runCheck function in scripts/preflight.mjs executes commands that reference files within the target repository provided via the --repo argument.
Boundary markers: None. The skill assumes the contents of the target repository are safe to execute/import for the purpose of checking dependencies.
Capability inventory: The skill has the capability to execute shell commands (spawnSync), read files (readFileSync), and perform file system checks (existsSync) across the host machine.
Sanitization: There is no sanitization or validation of the contents of the files imported from the target repository before execution.

dp-dev-bootstrap