The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for utilizing Bun's shell automation and child process management. It specifically guides the agent to use Bun.$ for shell-style commands and Bun.spawn() for advanced process control, including managing standard streams and environment variables.
[EXTERNAL_DOWNLOADS]: The skill covers standard package management workflows using bun install and bun add to fetch dependencies from remote registries. It also references the use of the Web-standard fetch API for network operations.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it is designed to ingest and process untrusted project data, such as configuration files and source code.
Ingestion points: The skill instructs the agent to read package.json, bun.lock, bunfig.toml, and other project source files to understand the project context.
Capability inventory: The agent has access to powerful capabilities including command execution via Bun.$ and Bun.spawn(), file system modifications via Bun.write(), and network interactions via fetch() and Bun.serve().
Boundary markers: The SKILL.md file contains instructions to prioritize explicit user constraints and system instructions over the skill's own guidance.
Sanitization: The instructions in references/clis.md and references/core-apis.md explicitly warn the agent to avoid shell interpolation bugs and to be careful with user-provided shell input by using documented escaping patterns.

bun