openspdd-workflow

Transforms OpenSpec proposals into structured REASONS Canvas prompts for deterministic COBOL code generation.

6|Updated Feb 25, 2026
One-click install
npx skills add https://github.com/emilianoeloi/cowboy --skill openspdd-workflow-emilianoeloi
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: openspdd-workflow
Source: https://github.com/emilianoeloi/cowboy/tree/main/.github/skills/openspdd-workflow
Command: npx skills add https://github.com/emilianoeloi/cowboy --skill openspdd-workflow-emilianoeloi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Generating COBOL code directly with AI often produces formatting hallucinations, since COBOL enforces strict column rules (columns 7, 8-11, 12-72) that break compilation when violated. This Skill bridges architectural specification and code generation through a three-stage Spec-Driven Prompt Development workflow that eliminates impulsive, unplanned code generation. ## Core Features & Use Cases - Strategic Analysis (spdd-analysis): Reads OpenSpec change artifacts (proposal.md, design.md, tasks.md) and produces an analysis covering the four COBOL divisions, Working-Storage variables, PIC clauses, and risks like column limits and ON SIZE ERROR handling. - REASONS Canvas Generation (spdd-reasons-canvas): Converts the analysis into a rigid 7-dimension executable contract (Requirements, Entities, Approach, Structure, Operations, Norms, Safeguards) with a unique traceability ID linking spec to code. - Deterministic Generation (spdd-generate): Executes Canvas operations line-by-line with phase gating, then runs build, fitness, and smoke-test sensors before validating the OpenSpec change. - Use Case: When adding a new operation (e.g., simple interest calculation) to a GnuCOBOL calculator, run the three operations in sequence so the coder agent follows a pre-planned contract instead of improvising COBOL syntax. ## Quick Start Ask the agent to run spdd-analysis on your OpenSpec change proposal, then generate the REASONS Canvas, and finally execute spdd-generate to produce validated COBOL code.

Frequently Asked Questions about openspdd-workflow

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate COBOL code with AI without formatting errors?▼

Use a spec-driven prompt workflow that first analyzes the proposal, then builds a structured canvas defining exact column rules (Area A columns 8-11, Area B columns 12-72), and finally executes generation step-by-step. This prevents the column misalignment hallucinations common in direct AI COBOL generation.

What is a REASONS Canvas in prompt engineering?▼

A REASONS Canvas is a structured prompt contract with seven dimensions: Requirements, Entities, Approach, Structure, Operations, Norms, and Safeguards. It converts strategic analysis into an executable specification that a coding agent follows deterministically without replanning.

How does spec-driven development work with OpenSpec?▼

OpenSpec changes contain proposal.md, design.md, and tasks.md artifacts that an analysis operation reads to inventory COBOL divisions, variables, and risks. The resulting canvas links back to the spec via a unique ID, and tasks.md checkboxes are updated and validated with openspec validate --strict after generation.

Does this workflow support GnuCOBOL compilation checks?▼

Yes, the generation operation runs make build using cobc -x as a compilation gate requiring exit code 0. It also runs fitness function tests and a smoke test with simulated input before marking the change complete.

When should I not use a multi-stage prompt workflow for code generation?▼

For trivial one-line changes or languages without strict formatting rules, the three-stage analysis-canvas-generate cycle adds unnecessary overhead. It is designed for format-sensitive languages like COBOL where column errors break compilation and traceability to specifications matters.