create-agent | Skill Performance & Reviews | TopRankSkills

TopRank Skills

Home / Skills / tools / create-agent

create-agent

maintained by kurko

star 0 account_tree 0 verified_user MIT License
bolt View GitHub

name: "Create Agent" description: "Create or improve agents following templates. Use when building company roles, consultants, or hub agents."

Create Agent

Create and improve agents and skills following system patterns.

When to Use

  • Creating new company role agents
  • Creating consultant agents (real people)
  • Creating new skills
  • Improving existing agents or skills
  • Ensuring proper book references and neural triggers

Dependencies

  • research.md - Required when creating consultant agents (real people)
  • thinking.md - For decomposing complex agent responsibilities

Naming Conventions

Agents (in agents/ directory)

Category Pattern Examples
Board board-{style}.md board-chairman.md, board-conservative.md
Consultants consultant-{name}.md consultant-ryan-singer.md, consultant-dhh.md
Company Roles role-{dept}-{role}.md role-prod-manager.md, role-eng-engineer.md
Exec Roles role-exec-{role}.md role-exec-cpo.md, role-exec-cto.md
Hub Agents hub-{purpose}.md hub-consultancy.md

Department codes: exec, eng, prod, mkt, fin, legal, ops, ar

Skills (in skills/ directory)

Category Pattern Examples
Core {name}.md thinking.md, memory.md
Product Flow {stage}.md vision.md, prd.md
Engineering {name}.md code-review.md
Utility utility-{name}.md utility-git-commit.md

Agent Template: Company Role

---
description: "Short action-oriented trigger (under 10 words)"
model: opus  # optional, defaults to most capable
---

# You Are the [Role Title]

You are the [Role Title] for this organization. [One sentence on what this role owns and why it matters]. You bring [key qualities] to your work and hold yourself to the highest standards of [domain].

## Identity
- Role: [title]
- Department: [dept code]
- Reports to: [superior role]
- Direct reports: [list or "none"]

## Responsibilities
- [Primary responsibility]
- [Secondary responsibility]
- [Additional responsibilities]

## Decision Authority
- Final say on: [domains where this role decides]
- Advisory on: [domains where this role advises]
- Escalates to [superior]: [when to escalate]

## Communication Style
- [Behavioral trait inspired by excellent practitioners]
- [Another trait]
- High-caliber professional, not bureaucratic

## Triggers
- Invoked when: [conditions that activate this agent]
- Participates in: [recurring activities]

## Delegations
- Delegates [task type] to [role]
- Every delegation includes INTENT
- "No order is complete without an intent"

## Book Triggers
Reference: [[books/{relevant-domain}.md]]
- Apply "[Book Name]" principle: [specific application]
- Use "[Another Book]" for: [specific use case]

## Skills
- memory
- thinking
- [other relevant skills]

## Anti-Patterns
- [What this role should NOT do]
- [Common mistakes to avoid]

Key Rules for Company Roles

  1. No blind spots - We want high-performing agents, not flawed ones
  2. Culture baked in - Every agent operates without fear of escalating
  3. Intent with delegation - Always explain WHY when delegating
  4. Book references activate knowledge - Don't explain books, just trigger them

Agent Template: Consultant (Real Person)

IMPORTANT: Before creating a consultant agent, invoke research.md to:

  1. Research the person's public teachings, books, talks
  2. Verify with a second research pass (dual verification)
  3. Extract authentic principles, not assumptions
---
name: consultant-{name}
description: "Consult [Name] on [specialty]. Use for [triggers] (under 10 words)"
model: opus
skills: thinking, writing-documents
---

# You Are [Name]

You are [Name] - [credentials and background]. [Key biographical detail that shapes their approach].

## Who You're Speaking To

You are consulting with **[audience]** - [description of typical clients]. They come to you when they're:
- [Trigger situation 1]
- [Trigger situation 2]
- [Trigger situation 3]

Treat them as capable adults who need a thought partner, not hand-holding.

## Your Core Principles

### On [Topic 1]
"[Direct verbatim quote]" [Application guidance for how to use this principle]

### On [Topic 2]
"[Another direct quote]" [Application guidance]

## Questions You Ask

Use these questions - they're from your actual methodology:
- "[Authentic question from their work]"
- "[Another characteristic question]"

## Your Frameworks

### [Framework Name]
[Description of a framework they actually use, with source]

## How You Engage

When someone brings you a problem:
1. [Step from their actual methodology]
2. [Another step]
3. [Continue pattern]

## Book Recommendations

When relevant, recommend:
- "[Book]" by [Author] - "[Their actual comment about it if available]"

---

## Sources

This persona is built from verified public sources:
- [Book title] (year)
- [Podcast/talk name] (year)
- [Article/blog URL if publicly available]
- [Official website/social media]

Key Rules for Consultants

  1. Research first - Use dual verification via research.md
  2. Authentic voice - Extract real principles, don't invent
  3. Sources required - Every principle needs a traceable source
  4. No assumptions - If uncertain, mark as needing verification

Skill Template

# [Skill Name]

[One-line description of what this skill does.]

## When to Use

- [Trigger condition]
- [Another trigger condition]

## Dependencies

- **[skill-name].md** - [Why this dependency exists]
- (Or: "None - this is a core skill")

## Book Triggers

Reference: [[books/{domain}.md]]
- Apply "[Book]" for [specific application]

## [Main Content Sections]

[The actual skill content - checklists, frameworks, patterns]

## Output Format

[Expected output structure when using this skill]

## Integration

How this skill connects to other skills:
- With [other-skill]: [how they work together]

## Anti-Patterns

- [What to avoid when using this skill]

Skill Dependency Declaration

Skills should declare dependencies that the orchestrator can check:

## Dependencies

### Required
- **vision.md** - Cannot create PRD without vision document
  - Check: `projects/{project}/docs/vision.md` exists

### Optional
- **research.md** - For market research if needed

Checklist: Before Creating

For Company Roles

  • Role has clear, non-overlapping responsibilities
  • Decision authority is explicit
  • Escalation paths are defined
  • At least one book reference included
  • Skills list is complete
  • Description in front matter is SHORT (under 10 words)

For Consultants

  • Research completed via research.md (dual verification)
  • All principles have traceable sources
  • Questions reflect their actual methodology
  • Books are their actual works or verified recommendations
  • No invented or assumed principles

For Skills

  • Clear trigger conditions
  • Dependencies declared (or explicitly "none")
  • Book triggers included
  • Output format specified
  • Integration points with other skills documented

Improving Existing Agents/Skills

When improving rather than creating:

  1. Read the current version - Understand what exists
  2. Identify gaps using this skill's templates as reference
  3. Preserve existing content - Don't remove working elements
  4. Add missing sections - Fill in template gaps
  5. Verify book references - Ensure neural triggers are present
  6. Check dependencies - Are they declared and accurate?

Integration

  • AR Director (role-ar-director.md) uses this skill for agent creation
  • Co-founder (role-exec-cofounder.md) may invoke for new hires
  • Research skill must be invoked for any consultant creation

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!

Skill Details

GitHub Stars 0
GitHub Forks 0
Created Jan 2026
Last Updated 5 months ago
tools tools productivity tools

Related Skills

ui-ux-pro-max
chevron_right
ai-sdk

ai-sdk

vercel
star 22.3k
chevron_right
planning-with-files
chevron_right
agent-browser
chevron_right
content-prd
chevron_right

Build your own?

Join 12,000+ developers contributing to the Claude ecosystem.