The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides commands to download and execute installation scripts from Bun's official domain.
Example: curl -fsSLo "$tmpdir/bun-install.sh" https://bun.sh/install followed by bash "$tmpdir/bun-install.sh".
The instructions include a safety step to review the script content (cat "$tmpdir/bun-install.sh") before execution.
Similar patterns are provided for Windows using PowerShell with Invoke-WebRequest and Get-Content for manual review.
[COMMAND_EXECUTION]: Instructs the agent on how to use shell commands for environment setup and project management.
Commands include brew install, npm install -g, and various bun CLI operations for running, testing, and building code.
Uses powershell -ExecutionPolicy Bypass for the Windows installation process.
[EXTERNAL_DOWNLOADS]: Facilitates downloading software and dependencies from external sources.
Downloads the Bun runtime from the official website (bun.sh).
References standard package registries and GitHub repositories for project dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill describes processes that involve reading and acting upon external project files, which constitutes a vulnerability surface.
Ingestion points: Project initialization (bun init), template creation (bun create), and reading environment variables from .env files in SKILL.md.
Boundary markers: Absent; there are no specific delimiters defined to prevent the agent from following instructions embedded in project files or data.
Capability inventory: The skill demonstrates file system access (Bun.file, Bun.write), network services (Bun.serve), and command execution (bun run, bunx).
Sanitization: No specific sanitization or validation logic is described for handling content within ingested project files or environment variables.

bun-development