mcpField
S
skill-archetypes

skill-archetypes

github.com/majesticlabs-devAgent workflows5046 installsUpdated 7w ago
Repository

About skill-archetypes

A reference skill for choosing a structure suited to a task, covering CLI, methodology, safety, and orchestration skills. It also describes reusable prompt, checklist, and validation-loop patterns.

Key features

  1. 01Four skill archetypes
  2. 02Structure guidance by task type
  3. 03Exact-prompt handoff pattern
  4. 04Checklist and validation patterns

Use cases

  1. 01Structuring a CLI reference skill
  2. 02Documenting a development methodology
  3. 03Adding approvals to risky operations
  4. 04Defining a multi-step agent workflow

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/majesticlabs-dev/majestic-marketplace --skill skill-archetypes

After installing, type/skill-archetypesin Claude Code to confirm it loaded.

Details

From the project README

Skill Archetypes

Identify which archetype fits best and follow its structure.

1. CLI Reference Skill

For tools, CLIs, APIs. Structure by operations, minimize prose.

## Authentication
[How to authenticate]

## Core Operations
| Command | Description |
|---------|-------------|
| `tool init` | Initialize project |
| `tool run` | Execute task |

## Common Workflows
[Task-oriented examples]

Examples: kamal-coder, wrangler-coder, gh-cli

2. Methodology Skill

For development practices, workflows, philosophies.

## Core Philosophy
[Why this approach matters - 2-3 sentences max]

## The Process
1. [Step with criteria]
2. [Step with criteria]

## Decision Criteria
| Situation | Action |
|-----------|--------|
| [condition] | [response] |

Examples: tdd-workflow, dhh-coder, founder-mode

3. Safety/Security Skill

For operations with risk. Include tiered approvals.

## Risk Tiers

| Tier | Operations | Approval |
|------|------------|----------|
| Low | Read-only queries | Auto |
| Medium | Modifications | Confirm |
| High | Destructive ops | Explicit |

## Blocking Patterns
[What to never do]

## Allowing Patterns
[Safe operations]

Examples: infra-security-review, devops-verifier

4. Orchestration Skill

For multi-step processes that coordinate other tools.

## Quick Start
[Minimal invocation]

## Workflow Phases
1. [Phase]: [What happens]
2. [Phase]: [What happens]

## Machine-Readable Output
[JSON/YAML schema for automation]

Examples: build-task-workflow, quality-gate

Advanced Patterns

THE EXACT PROMPT Pattern

For reproducible agent-to-agent handoffs, encode prompts in ALL CAPS:

## THE EXACT PROMPT

ANALYZE THE FOLLOWING CODE FOR SECURITY VULNERABILITIES:
1. CHECK FOR HARDCODED SECRETS
2. IDENTIFY SQL INJECTION RISKS
3. FLAG INSECURE DEPENDENCIES

RETURN FINDINGS AS JSON WITH SEVERITY LEVELS.
Checklist Pattern

For multi-step processes with progress tracking:

## Checklist

- [ ] Step 1: [Action with success criteria]
- [ ] Step 2: [Action with success criteria]
- [ ] Step 3: [Action with success criteria]

Mark each complete before proceeding.
Feedback Loop Pattern

For operations requiring validation:

## Validation Loop

1. Execute action
2. Verify result matches criteria
3. If FAIL: diagnose → fix → goto 1
4. If PASS: proceed to next step

Never advance without validation passing.

Comments

Markdown supported