From toby-essentials
Presents menu of 8 pre-built harness use cases for research/analysis, content creation, media/marketing, engineering; launches selected one via harness:harness skill.
npx claudepluginhub tobyilee/toby-plugins --plugin toby-essentialsThis skill uses the workspace's default tool permissions.
Launch a pre-built harness use case with a single selection. This skill acts as a quick-start
Configures harnesses by defining specialist agents and generating skills for domain-specific project automation. Use for initial setup, architecture design, expansion, audits, or maintenance.
Designs multi-agent harness architectures for long-running AI apps using GAN-inspired Generator-Evaluator pattern, Sprint Contract negotiation, and quality evaluation loops. For agent orchestration, context management, and complex full-stack planning-generation-evaluation cycles.
Guides harness engineering for AI agents: context/memory management, guardrails, AGENTS.md/CLAUDE.md repo instructions, evals, observability, and orchestration.
Share bugs, ideas, or general feedback.
Launch a pre-built harness use case with a single selection. This skill acts as a quick-start
menu for the harness:harness skill, removing the need to remember or type full prompt templates.
| # | Category | Use Case | Pattern | Description |
|---|---|---|---|---|
| 1 | Research & Analysis | Deep Research | Fan-out/Fan-in | Multi-angle investigation with cross-validation |
| 2 | Research & Analysis | Code Review | Fan-out/Fan-in | Parallel architecture/security/perf analysis |
| 3 | Content Creation | Website Development | Pipeline | Full-stack from wireframe to deployment |
| 4 | Content Creation | Webtoon Production | Producer-Reviewer | Story, design, layout with peer review |
| 5 | Media & Marketing | YouTube Content | Supervisor | Trend research, scripting, SEO optimization |
| 6 | Media & Marketing | Marketing Campaign | Producer-Reviewer | Market research, copy, visuals, A/B testing |
| 7 | Engineering | Tech Documentation | Pipeline | API docs from codebase analysis |
| 8 | Engineering | Data Pipeline | Hierarchical | Schema, ETL, validation, monitoring |
First, print the full menu as a formatted text block so the user can see every option at a glance:
π§ Harness Use Cases
Research & Analysis
1. Deep Research β Multi-angle investigation with cross-validation (Fan-out/Fan-in)
2. Code Review β Parallel architecture/security/perf analysis (Fan-out/Fan-in)
Content Creation
3. Website Development β Full-stack pipeline from wireframe to deployment (Pipeline)
4. Webtoon Production β Story, design, layout with peer review (Producer-Reviewer)
Media & Marketing
5. YouTube Content β Trend research, scripting, SEO optimization (Supervisor)
6. Marketing Campaign β Market research, copy, visuals, A/B testing (Producer-Reviewer)
Engineering
7. Tech Documentation β API docs from codebase analysis (Pipeline)
8. Data Pipeline β Schema, ETL, validation, monitoring (Hierarchical)
After displaying the full menu, use AskUserQuestion with 4 category options. Each category label includes its two use case numbers so the user knows what's inside:
If the user picks "Other", ask them to describe their custom use case as free text, then pass that directly to the harness skill.
After the user picks a category, use AskUserQuestion again with the 2 specific use cases in that category. This gives a clean two-step selection: category β use case.
If the category has only one match (user already named a specific use case in "Other"), skip this step.
Read the selected use case's full prompt from references/use-cases.md.
The file contains numbered sections (## 1. Deep Research, etc.) with the exact prompt
in a fenced code block.
Before launching, show the user the selected prompt and ask if they want to customize it. If the user provides additional context (e.g., "but use Vue instead of React"), append that to the base prompt.
Invoke the harness:harness skill with the final prompt. Pass the prompt as the skill argument:
Skill: harness:harness
Args: <the composed prompt>
The harness skill handles the full 6-phase workflow: domain analysis, team architecture, agent generation, skill generation, integration, and validation.
For context when the user asks about patterns: