skills.rest
onesmash
onesmashCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

workflow-creator

Generates durable workflow scaffolds with contracts, policies, verifiers, and prompts from a spec.json blueprint.

Updated May 13, 2026
Search Tags:#automation#workflow#spec-driven#code generation#scaffolding#durable execution#verifiers
One-click install
npx skills add https://github.com/onesmash/slm-as-harness --skill workflow-creator-onesmash
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: workflow-creator
Source: https://github.com/onesmash/slm-as-harness/tree/main/skills/durable-workflow-runtime/workflow-creator
Command: npx skills add https://github.com/onesmash/slm-as-harness --skill workflow-creator-onesmash

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve? Creating a new durable workflow by hand means writing stage contracts, runtime state, transition policy, GraphBuilder plumbing, prompt assets, verifiers, tests, and catalog entries consistently. This Skill automates that authoring work by generating the full workflow package from a single workflow-local spec.json blueprint. ## Core Features & Use Cases - Spec-Driven Generation: Create a workflow scaffold, fill in workflow-runtime/workflows/<workflow_id>/spec.json with business stages, then regenerate contracts, state, policy, prompts, verifiers, flowcharts, and regression tests with --force. - Routing and Verification DSL: Declare transitions, outcome routes, repair conditions, verifier rules, verifier templates, and custom verifier requirements that the generator turns into concrete policy.py and verifiers.py code. - Binding and Shortcut Registration: Register the workflow in workflow-binding.json and create a slash-only shortcut skill so Claude Code can discover /<workflow_id>. - Use Case: A user describes a PDF processing pipeline with intake, extraction, review, and summary stages; the Skill scaffolds the workflow, generates all runtime files, and produces an agent-review checklist before shipping. ## Quick Start Ask the AI to create a new durable workflow with a chosen kebab-case workflow id and a short flow description, then fill in the generated spec.json and rerun the creator with --force.

Frequently Asked Questions about workflow-creator

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

FAQPage Schema
How do I create a new durable workflow with workflow-creator?▼

Run create_workflow.py with --workflow-id and --flow-description to create the scaffold and workflow-local spec.json. Fill spec.json with concrete business stages, then rerun the script with --force to regenerate contracts, policy, prompts, verifiers, and tests.

What is the spec.json file in a generated workflow?▼

spec.json is the durable blueprint stored at workflow-runtime/workflows/<workflow_id>/spec.json. It declares stages, prompts, output schemas, transitions, verifier rules, dependencies, and regression tests, and it is the only long-lived spec you should maintain.

Should new workflow ids use kebab-case or snake_case?▼

New workflows should use hyphen-separated kebab-case ids such as pdf-processing. Snake_case ids remain accepted so existing workflows can be regenerated with --force, and the creator derives the underscore Python module name automatically.

When should I use verifier templates versus custom verifier requirements?▼

Use verifier_templates when built-in templates like min_count, conditional_required, or artifact_list_policy express the invariant cleanly. Use custom_verifier_requirements for domain-specific acceptance logic, which generates requirement-scoped verifiers.py scaffolds to complete before review.

Why does workflow creation require a subagent review pass?▼

The generator owns only mechanical surfaces like contract.py and policy.py, so semantic correctness needs separate review. The generated references/agent-review.md checklist drives subagent passes over prompts, verifiers, contracts, and runtime flow, and the Skill blocks until the user authorizes subagent use.

Can workflow-creator start or resume workflow execution?▼

No. It is an authoring surface only and explicitly must not call bridge.py start, resume, or preflight. It creates and regenerates workflow files, binding entries, and shortcut skills, but never mutates runtime run state.

Related Skills

Explore More Software Engineering Skills→
rickvangrickvang

playbook-composer

Compose reusable multi-stage operating models with stages, handoffs, decision rights, and quality gates.

Community
Advanced
p-sree-sai-pavanp-sree-sai-pavan
5

prose

Coordinate multi-agent workflows with durable state across filesystems or databases.

Community
Advanced
gganbukim1gganbukim1

ai-first-engineering

Design engineering processes and architecture for teams shipping AI-generated implementation output.

Community
Intermediate
BlocUnited-LLCBlocUnited-LLC
14

runtime-change

Review or implement Mozaiks OSS runtime and platform host changes.

Official
Advanced