Generates custom sub-agent directories with agent.md prompts, rules, and skills.
npx skills add https://github.com/HACK-WU/skills --skill create-sub-agent-hack-wu
Please help me install this Agent Skill. Skill: create-sub-agent Source: https://github.com/HACK-WU/skills/tree/main/skills/create-sub-agent Command: npx skills add https://github.com/HACK-WU/skills --skill create-sub-agent-hack-wu
agents/{agent-name}/ structure with agent.md plus optional rules/ and skills/ subdirectories. - Structured Prompt Authoring: Writes agent.md following a fixed template covering role definition, capability assumptions, three-tier file permissions, task boundaries, input/output contracts, and failure reporting. - Validation Checklist: Self-checks the output against a verification list including the three-part output contract and the no-redelegation constraint. - Use Case: When you need a dedicated API testing agent, ask to create an api-tester sub-agent and receive a complete agent.md with permission boundaries, a three-section reporting contract, and a companion skill for HTTP assertions. ## Quick Start Ask the AI to create a sub-agent named api-tester that validates HTTP API responses and reports results in the three-section format.High-intent search queries and answers about installing and using this skill.
Define the agent name in kebab-case, its role, task boundaries, and delegation triggers, then generate an agents/{name}/ directory containing an agent.md prompt file. The agent.md follows a fixed template with role definition, capability assumptions, permission tiers, and input/output contracts.
An agent.md needs frontmatter with description and provider fields, plus sections for role positioning, capability assumptions, three-tier file permissions (writable, needs-confirmation, forbidden), task boundaries, a five-field input contract, execution requirements, a three-section output contract, failure reporting, and a self-check list.
Skip creating a new agent when an existing agent already covers the functionality, when a one-off task can be delegated to a generic sub-agent, or when you only need behavioral constraints, which belong in rules files instead of a dedicated agent.
Other skills reference the agent with use_agent({agent-name}) syntax in their SKILL.md, without copying file paths or agent content. The agent.md in agents/{agent-name}/ remains the single source of truth, and delegation passes the five-field input contract.
No. The template explicitly forbids re-delegation to prevent uncontrolled delegation chains and unclear responsibility. A sub-agent must refuse tasks outside its boundary and report back to the main agent instead.
brolagGenerate structured delegation specs for Codex or async agents.
friday-platformDefine delegate() call parameters for spawning child agents with inherited context.
ai-ecoverseCreate, edit, and validate SLICC agent skills with frontmatter and companion files.
ryancnelsonDelegate Jira issue queries to a specialized sub-agent.