agentleFS
YParesYPares/agent-skills/nushell-plugin-builder/SKILL.md

nushell-plugin-builder

Skill30 starsChanged 10 months agoMITView on GitHub

Guide for creating Nushell plugins in Rust using nu_plugin and nu_protocol crates. Use when users want to build custom Nushell commands, extend Nushell with new functionality, create data transformations, or integrate external tools/APIs into Nushell. Covers project setup, command implementation, streaming data, custom values, and testing.

  • Installs packages

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 4 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use for: transformations, simple filters, single value operations

    line 100
  2. Distinctive:

    Use for: streaming transformations, lazy processing, large datasets

    line 106
  3. Very specific:

    Always return LabeledError with span information:

    line 147
  4. Distinctive:

    Use for production plugins

    line 159
  5. Very specific:

    Check engine.is_using_stdio() before attempting stdio access

    line 291
  6. Very specific:

    Always use paths relative to engine.get_current_dir()

    line 294
  7. Distinctive:

    Never assume current working directory

    line 295
  8. Very specific:

    Use scripts/init_plugin.py to scaffold a new plugin with proper structure:

    line 318

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Nushell Plugin Builder

Overview

This skill helps create Nushell plugins in Rust. Plugins are standalone executables that extend Nushell with custom commands, data transformations, and integrations.

Quick Start

1. Create Plugin Project

cargo new nu_plugin_<name>
cd nu_plugin_<name>
cargo add nu-plugin nu-protocol

2. Basic Plugin Structure

use nu_plugin::{EvaluatedCall, MsgPackSerializer, serve_plugin};
use nu_plugin::{EngineInterface, Plugin, PluginCommand, SimplePluginCommand};
use nu_protocol::{LabeledError, Signature, Type, Value};

struct MyPlugin;

impl Plugin for MyPlugin {
    fn version(&self) -> String {
        env!("CARGO_PKG_VERSION").into()
    }

    fn commands(&self) -> Vec<Box<dyn PluginCommand<Plugin = Self>>> {
        vec![Box::new(MyCommand)]
    }
}

struct MyCommand;

impl SimplePluginCommand for MyCommand {
    type Plugin = MyPlugin;

    fn name(&self) -> &str {
        "my-command"
    }

    fn signature(&self) -> Signature {
        Signature::build("my-command")
            .input_output_type(Type::String, Type::Int)
    }

    fn run(
        &self,
        _plugin: &MyPlugin,
        _engine: &EngineInterface,
        call: &EvaluatedCall,
        input: &Value,
    ) -> Result<Value, LabeledError> {
        match input {
            Value::String { val, .. } => {
                Ok(Value::int(val.len() as i64, call.head))
            }
            _ => Err(LabeledError::new("Expected string input")
                .with_label("requires string", call.head))
        }
    }
}

fn main() {
    serve_plugin(&MyPlugin, MsgPackSerializer)
}

3. Build and Install

# Build
cargo build --release

# Install to cargo bin
cargo install --path . --locked

# Register with nushell
plugin add ~/.cargo/bin/nu_plugin_<name>  # Add .exe on Windows
plugin use <name>

# Test
"hello" | my-command

Command Types

SimplePluginCommand

For commands that operate on single values:

  • Input: &Value
  • Output: Result<Value, LabeledError>
  • Use for: transformations, simple filters, single value operations

PluginCommand

For commands that handle streams:

  • Input: PipelineData
  • Output: Result<PipelineData, LabeledError>
  • Use for: streaming transformations, lazy processing, large datasets

See references/advanced-features.md for streaming examples.

Defining Command Signatures

Input-Output Types

use nu_protocol::{Signature, Type};

Signature::build("my-command")
    .input_output_type(Type::String, Type::Int)

Common types: String, Int, Float, Bool, List(Box<Type>), Record(...), Any

Parameters

Signature::build("my-command")
    // Named flags
    .named("output", SyntaxShape::Filepath, "output file", Some('o'))
    .switch("verbose", "enable verbose output", Some('v'))
    // Positional arguments
    .required("input", SyntaxShape::String, "input value")
    .optional("count", SyntaxShape::Int, "repeat count")
    .rest("files", SyntaxShape::Filepath, "files to process")

Accessing Arguments

fn run(&self, call: &EvaluatedCall, ...) -> Result<Value, LabeledError> {
    let output: Option<String> = call.get_flag("output")?;
    let verbose: bool = call.has_flag("verbose")?;
    let input: String = call.req(0)?;  // First positional
    let count: Option<i64> = call.opt(1)?;  // Second positional
    let files: Vec<String> = call.rest(2)?;  // Remaining args
}

Error Handling

Always return LabeledError with span information:

Err(LabeledError::new("Error message")
    .with_label("specific issue", call.head))

This shows users exactly where the error occurred in their command.

Serialization

MsgPackSerializer (Recommended)

  • Binary format, much faster
  • Use for production plugins

JsonSerializer

  • Text-based, human-readable
  • Useful for debugging

Choose in main():

serve_plugin(&MyPlugin, MsgPackSerializer)  // Production
// serve_plugin(&MyPlugin, JsonSerializer)  // Debug

Common Patterns

String Transformation

Value::String { val, .. } => {
    Ok(Value::string(val.to_uppercase(), call.head))
}

List Generation

let items = vec![
    Value::string("a", call.head),
    Value::string("b", call.head),
];
Ok(Value::list(items, call.head))

Record (Table Row)

use nu_protocol::record;

Ok(Value::record(
    record! {
        "name" => Value::string("example", call.head),
        "size" => Value::int(42, call.head),
    },
    call.head,
))

Table (List of Records)

let records = vec![
    Value::record(record! { "name" => Value::string("a", span) }, span),
    Value::record(record! { "name" => Value::string("b", span) }, span),
];
Ok(Value::list(records, call.head))

See references/examples.md for complete working examples including:

  • Filtering streams
  • HTTP API calls
  • File system operations
  • Multi-command plugins

Development Workflow

Iterative Development

# Build
cargo build

# Test (debug build)
plugin add target/debug/nu_plugin_<name>
plugin use <name>
"test" | my-command

# After changes, reload
plugin rm <name>
plugin add target/debug/nu_plugin_<name>
plugin use <name>

Automated Testing

[dev-dependencies]
nu-plugin-test-support = "0.109.1"
#[cfg(test)]
mod tests {
    use nu_plugin_test_support::PluginTest;

    #[test]
    fn test_command() -> Result<(), nu_protocol::ShellError> {
        PluginTest::new("myplugin", MyPlugin.into())?
            .test_examples(&MyCommand)
    }
}

See references/testing-debugging.md for debugging techniques and troubleshooting.

Advanced Features

Streaming Data

For lazy processing of large datasets, use PipelineData:

impl PluginCommand for MyCommand {
    fn run(&self, input: PipelineData, ...) -> Result<PipelineData, LabeledError> {
        let filtered = input.into_iter().filter(|v| /* condition */);
        Ok(PipelineData::ListStream(ListStream::new(filtered, span, None), None))
    }
}

Engine Interaction

// Get environment variables
let home = engine.get_env_var("HOME")?;

// Set environment variables (before response)
engine.add_env_var("MY_VAR", Value::string("value", span))?;

// Get plugin config from $env.config.plugins.<name>
let config = engine.get_plugin_config()?;

// Get current directory for path resolution
let cwd = engine.get_current_dir()?;

Custom Values

Define custom data types that extend beyond Nushell's built-in types. See references/advanced-features.md for complete guide.

Important Constraints

Stdio Restrictions

  • Plugins cannot use stdin/stdout (reserved for protocol)
  • Check engine.is_using_stdio() before attempting stdio access

Path Handling

  • Always use paths relative to engine.get_current_dir()
  • Never assume current working directory

Version Compatibility

  • Match nu-plugin and nu-protocol versions
  • Both should match target Nushell version

Reference Documentation

  • references/plugin-protocol.md - Protocol details, serialization, lifecycle
  • references/advanced-features.md - Streaming, EngineInterface, custom values
  • references/examples.md - Complete working examples and patterns
  • references/testing-debugging.md - Development workflow, debugging, troubleshooting

External Resources

Template Script

Use scripts/init_plugin.py to scaffold a new plugin with proper structure:

python3 scripts/init_plugin.py <plugin-name> [--output-dir <path>]

This creates a complete working plugin template ready to customize.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.