Skip to content
Back to skills

Create Subagent

ASecurity

"创建自定义子代理。为专门任务定制子代理(代码审查、调试、领域助手等)。"disable-model-invocation: true

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 16, 2026
ai-agentsdocumentation

Works with

  • cursor

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned October 16, 2026

npx -y skills add 0mlkrizzz655335v/claude-skills-zh --skill create-subagent --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Create Subagent?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Create Subagent
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/0mlkrizzz655335v-create-subagent/badge)](https://www.skillsdirectory.com/skills/0mlkrizzz655335v-create-subagent)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: create-subagent
description: "创建自定义子代理。为专门任务定制子代理(代码审查、调试、领域助手等)。"disable-model-invocation: true
---


# Create Custom Subagents

Create the reusable agent requested by the user. Distinguish persistent agent configuration from launching a one-off collaborator in this conversation.

## Resolve the supported mechanism

Use the current agent catalog, runtime configuration, and official documentation to identify how the target host registers custom agents. Do not assume `.cursor/agents/*.md` is consumed by every Codex runtime. If the target supports that format, read [file-based agent format](references/file-based-agents.md); otherwise use its actual schema and supported model/permission fields.

For a temporary independent task, current collaboration tools may be sufficient; follow their actual schema rather than inventing `subagent_type` or model IDs. A temporary spawned agent does not prove persistent registration.

## Author and validate

Infer the task, scope, and reusable knowledge from the request. Keep the description precise: name the tasks that need this specialist instead of adding “use proactively” or “after every code edit” to all agents. Preserve explicit model choices and do not broaden permissions.

Write only the task-specific prompt, necessary output expectations, and boundaries that change behavior. Honor the requested project or personal location after resolving the runtime convention. Avoid installing unrelated example agents.

Verify configuration syntax and references, then confirm runtime discovery when possible. A small independent invocation is useful when it can meaningfully test the agent within authorized scope. Report separately what was written, what loaded, and what actually ran. If persistence is unavailable, preserve the prepared prompt and identify that specific limitation without claiming installation succeeded.

Files in this skill

  • SKILL.md1.9 KB
  • references/file-based-agents.md1.2 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.