Code Execution Skills

Skills for running code and shell commands via artifacts and execute_shell in the Daytona sandbox

/api/skills/code

3 Skills

execute-code

How to run Python code safely using artifacts and execute_shell in the Daytona sandbox.

Skill python
Content Type
Static (Markdown)
Language
python

Trigger Keywords

codescriptexecutesandboxpythonartifacts

Instructions Preview

# Code Execution Skill Use this skill whenever you need to run custom code in the Daytona sandbox. This skill **replaces the legacy `execute_code` tool**. All code execution now happens via: - scripts stored as artifacts under `scripts/`, and - the `execute_shell` tool, using its `command` field to run those scripts. The **canonical pattern** is: 1. **Write scripts as artifacts** under `scripts/` (via `write_artifact`). 2. **Write inputs as artifacts** under `files/` (via `write_artifact`) or pull them from Vertesia with `fetch_document`. 3. **Run scripts via `execute_shell`** in a bash shell by setting the `command` field (for example, `python /home/daytona/scripts/analyze.py`). 4. **Prefer writing outputs to `/home/daytona/out/`** so they become `out/*` artifacts and can be reused across steps instead of returning large raw text in tool results. Avoid relying on ad-hoc inline code execution where possible; scripts in artifacts are easier to inspect, patch, and reuse. --- ## 1...

shell-automation

Run bash commands and small shell pipelines safely in the Daytona sandbox using execute_shell.

Skill python
Content Type
Static (Markdown)
Language
python

Trigger Keywords

shellbashcliautomationsandboxcommands

Instructions Preview

# Shell Automation Skill Use this skill when you need to run shell commands (bash) in the Daytona sandbox: - Inspect files and directories. - Run small CLIs and utilities. - Glue together scripts and data with simple pipelines. All execution goes through the `execute_shell` tool; this skill explains how to use it effectively. ## 1. Explore the sandbox Basic commands you can run via `execute_shell.command`: - `pwd` – show current directory (`/home/daytona`). - `ls -R` – list files and subdirectories. - `cat`, `head`, `tail` – inspect files. - `grep` / `rg` – search in files. Example: ```bash ls -la /home/daytona ls -la /home/daytona/scripts ls -la /home/daytona/files ``` ## 2. Combine with artifacts Artifacts map to sandbox directories: - `scripts/*` → `/home/daytona/scripts/` - `files/*` → `/home/daytona/files/` - `docs/*` → `/home/daytona/documents/` (snapshots of original documents) - `out/*` → `/home/daytona/out/` Typical pattern: 1. Use `write_artifact` to create files...

node-scripting

Deprecated; prefer the execute-code and shell-automation skills, which use Python and shell commands in the Daytona sandbox.

Skill python
Content Type
Static (Markdown)
Language
python

Trigger Keywords

deprecatedscripting

Instructions Preview

# Deprecated Scripting Skill This skill is deprecated and kept only for backward compatibility. For any scripting or automation needs in the Daytona sandbox: - Use the **execute-code** skill for Python scripts under `scripts/`. - Use the **shell-automation** skill for shell pipelines and orchestration. Write your logic in Python scripts stored as artifacts and execute them with `execute_shell`, and direct any reusable outputs to `/home/daytona/out/` so they sync back as `out/*` artifacts.