promptcmd

Appearance

Frontmatter Configuration ​

The YAML frontmatter defines how your prompt behaves and what parameters it accepts.

yaml
---
# Frontmatter: YAML configuration
model: model_name
input:
  schema:
    param1: type, description
    param2?: type, optional description
---
Your prompt text here with {{param1}} and {{param2}}.

Model Specification ​

The model field determines which AI provider and model to use:

yaml
model: anthropic/claude-sonnet-4

Options include:

  • Provider/Model format, e.g., anthropic/claude-sonnet-4, openai/gpt-4o, google/gemini-pro
  • Provider only, e.g., anthropic, openai, google. The default set model for the selected provider will be used
  • Variant name: Custom configurations you've defined (e.g., rust-coder, quick-responder)
  • Group name: Load balancing groups (e.g., balanced, weighted)

See Model Resolution for how model names are resolved.

Input Schema ​

The input schema defines the parameters your prompt command accepts. It controls CLI argument parsing, type validation, and help text generation.

Basic Schema Syntax ​

Schemas are defined in the frontmatter's input.schema section:

yaml
---
input:
  schema:
    name: string, User name
    age: integer, User age in years
    verbose?: boolean, Enable detailed output
---

Supported Types ​

String ​

yaml
---
input:
  schema:
    message: string, Message to process
    author?: string, Optional author name
---
bash
command --message "Hello, world!" --author "Alice"

Boolean ​

yaml
---
input:
  schema:
    summarize?: boolean, Summarize the text
    simplify?: boolean, Simplify the text
---
bash
command --summarize        # summarize = true
command --simplify         # simplify = true
command                    # both = false

Integer ​

yaml
---
input:
  schema:
    count: integer, Number of items
---
bash
command --count 5

Number ​

yaml
---
input:
  schema:
    threshold?: number, Confidence threshold
---
bash
command --threshold 0.95

Enum (Choices) ​

yaml
---
input:
  schema:
    level(enum, Log Level)?: [debug, info, warn, error] #optional
    format(enum, How to format the output): [json, yaml, toml]
---
bash
command --level info --format json

Help output shows available choices:

--format <format>  How to format the output [possible values: json, yaml, toml]
--level <level>    Log Level [possible values: debug, info, warn, error]

Accepting Standard Input (stdin) ​

Standard Input is accessible directly in the template without explicit declaration in the input schema. Use the stdin helper to retrieve it:

handlebars
Summarize the following text: {{stdin}}

Output Schema ​

For models supporting structured output, you can describe the schema of the output in the Frontmatter.

WARNING

Structured output is tested and working with google models. Support for others will be added.

yaml
---
model: google
output:
  format: json
  schema:
    person_name: string
    age: integer
    email: string
    occupation: string
---
Extract the specified information from the text below and return it in JSON
format based on the supplied schema.

Sarah Johnson is a 28-year-old software engineer currently working at TechCorp.
She has expertise in Python, JavaScript, and cloud architecture. You can reach
her at sarah.j@email.com"
json
{
  "email": "sarah.j@email.com",
  "age": 28,
  "occupation": "software engineer",
  "person_name": "Sarah Johnson"
}