agent-creator

Guide for creating custom Claude Code subagents. Use when user wants to create a new agent (or update an existing agent) that handles specific types of tasks with custom prompts, tool restrictions, and permissions. Triggers on requests to create agents, subagents, custom agents, or when user wants specialized AI assistants for specific workflows.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "agent-creator" with this command: npx skills add walletconnect/skills/walletconnect-skills-agent-creator

Agent Creator

Create subagents as Markdown files with YAML frontmatter. Store in:

  • ~/.claude/agents/ - User-level (all projects)
  • .claude/agents/ - Project-level (current project only)

Subagent File Structure

---
name: agent-name
description: When Claude should delegate to this agent
tools: Read, Grep, Glob
model: sonnet
---

System prompt instructions for the agent go here.
The agent receives only this prompt, not the full Claude Code system prompt.

Required Fields

FieldDescription
nameUnique identifier (lowercase, hyphens only)
descriptionWhen to delegate - Claude uses this to decide. Include "use proactively" for automatic delegation

Optional Fields

FieldDefaultDescription
toolsAllComma-separated tool names. See configuration.md for full list
disallowedToolsNoneTools to deny from inherited set
modelsonnetsonnet, opus, haiku, or inherit
permissionModedefaultdefault, acceptEdits, dontAsk, bypassPermissions, plan
skillsNoneSkills to load into agent context at startup
hooksNoneLifecycle hooks (PreToolUse, PostToolUse, Stop)

For detailed options and all available tools, see references/configuration.md.

System Prompt Guidelines

  • Be specific about what the agent should do when invoked
  • Include a clear workflow (numbered steps)
  • Define output format expectations
  • Keep focused on one domain/task

Examples

Read-Only Reviewer

---
name: code-reviewer
description: Reviews code for quality and security. Use proactively after code changes.
tools: Read, Grep, Glob, Bash
model: inherit
---

Review code and provide actionable feedback.

When invoked:
1. Run git diff to see recent changes
2. Focus on modified files
3. Check for: clarity, duplication, error handling, security, tests

Format feedback by priority:
- Critical (must fix)
- Warnings (should fix)
- Suggestions (consider)

Read-Write Fixer

---
name: debugger
description: Debug and fix errors, test failures, unexpected behavior. Use proactively for issues.
tools: Read, Edit, Bash, Grep, Glob
---

Debug issues with systematic root cause analysis.

When invoked:
1. Capture error message and stack trace
2. Identify reproduction steps
3. Isolate failure location
4. Implement minimal fix
5. Verify solution

Provide: root cause, evidence, code fix, prevention.

Hook-Validated Agent

---
name: db-reader
description: Execute read-only database queries for analysis and reports.
tools: Bash
hooks:
  PreToolUse:
    - matcher: "Bash"
      hooks:
        - type: command
          command: "./scripts/validate-readonly.sh"
---

Execute SELECT queries only. Explain that write access is unavailable if requested.

Best Practices

  1. Focus agents narrowly - One agent per domain/task
  2. Write clear descriptions - Claude uses these for delegation decisions
  3. Limit tools appropriately - Grant only what's needed
  4. Use hooks for validation - When you need finer control than tool restrictions
  5. Test with real tasks - Verify the agent behaves as expected

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Web3

skill-writing

No summary provided by upstream source.

Repository SourceNeeds Review
Web3

code-review

No summary provided by upstream source.

Repository SourceNeeds Review
Web3

command-creator

No summary provided by upstream source.

Repository SourceNeeds Review