Prompt4Py

Programmatic prompt template for Python.

Prompt Engineering Skill

Prompt4Py includes the prompt-engineering agent skill for designing, reviewing, and validating prompts as controlled reasoning interfaces.

For a non-trivial prompt task, the skill helps an agent:

• define one objective, clear evidence boundaries, and a machine-checkable output contract;
• separate instructions, constraints, reusable context, and runtime input;
• keep retries, persistence, acceptance, and other workflow responsibilities in code or an external harness;
• validate the prompt structure before delivery and report remaining unknowns honestly.

The goal is not more prompt text. It is a smaller, explicit decision surface with a verification path. Prompt4Py is the recommended Python implementation when a structured template is needed.

Why Prompt4Py?

Prompt4Py is inspired by two practical ideas:

• Prompts should be structured, programmable artifacts: define templates in Python and render them with runtime values.
• Prompt layout is part of the design. The placement of instructions, context, and few-shot examples can affect LLM performance, so templates keep these elements explicit and controllable.

Research on Demos' Position in Prompt (DPP) bias shows that moving demonstrations within a prompt can materially affect accuracy and output stability. See Cobbina and Zhou, Where to show Demos in Your Prompt: A Positional Bias of In-Context Learning (2025).

Installation

Python SDK

pip install -U prompt4py

Prompt Engineering Skill

Codex

$skill-installer install https://github.com/vortezwohl/Prompt4Py/tree/main/.skills/prompt-engineering

Restart Codex after installation.

Other Agent Skills-compatible tools

Copy .skills/prompt-engineering/ into the tool's supported skills location, then invoke:

prompt-engineering

Important

Review SKILL.md, its references, and its validation script before installing it. A skill is executable agent guidance.

Quick Start

1. Create a prompt template

   from prompt4py import GeneralTemplate
   
   # Create your prompt template
   prompt_template = GeneralTemplate()
   prompt_template.role = 'An NER machine'
   prompt_template.objective = 'Extract all {{entity_type}} from CONTEXT.'
   prompt_template.instruction = {
       1: 'Think deeply on every entities in CONTEXT',
       2: 'Extract all {{entity_type}}',
       3: 'Output the entities you have extracted'
   }
   prompt_template.constraint = 'Do not include any markdown grams'
   prompt_template.capability = 'Extract entities'
   prompt_template.context = '{{ent_1}}, {{ent_2}}, {{ent_3}}'
   prompt_template.output_dtype = 'str'
   prompt_template.output_format = 'jsonl'
   prompt_template.output_example = str([
       {
           'entity_type': '{{example_entity_type_1}}',
           'entity_text': '{{example_entity_text_1}}'
       }
   ])

2. Render the template

   # Render the template
   prompt = prompt_template.render(entity_type='PERSON', ent_1='John Lennon', ent_2='Joe Biden', ent_3='Charlemagne',
                                   example_entity_type_1='PERSON', example_entity_text_1='Elizabeth')
   print(prompt)

   the prompt would be rendered like this:

   ## _TIMESTAMP
   [82159.8475038]
   ## ROLE
   An NER machine
   ## OBJECTIVE
   Extract all PERSON from CONTEXT.
   ## INSTRUCTION
   - **1**: Think deeply on every entities in CONTEXT 
   - **2**: Extract all PERSON 
   - **3**: Output the entities you have extracted 
   ## CONSTRAINT
   Do not include any markdown grams
   ## CAPABILITY
   Extract entities
   ## CONTEXT
   John Lennon, Joe Biden, Charlemagne
   ## OUTPUT_DATATYPE
   str
   ## OUTPUT_FORMAT
   jsonl
   ## OUTPUT_EXAMPLE
   [{'entity_type': 'PERSON', 'entity_text': 'Elizabeth'}]

3. Invoke a chatbot / causal language model

   You would get response like below:

   {"entity_type": "PERSON", "entity_text": "John Lennon"}
   {"entity_type": "PERSON", "entity_text": "Joe Biden"}
   {"entity_type": "PERSON", "entity_text": "Charlemagne"}