菜单
Creates a new agent for this project. Use when the user asks to create, define, or build a new custom agent.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when the user wants to create, define, build, or package a new skill — including when they ask to turn a workflow into a reusable command, wrap domain knowledge into a slash-command, or describe a repeatable procedure they want Copilot to follow. Gathers name and purpose, infers a complete specification, writes a definition file, and invokes smaqit.L2 to compile a SKILL.md file.
Use when deploying a feature to an existing project's production environment through its already-proven CI/CD pipeline. Reads specs to identify the stack, target, and provisioning mode before any infrastructure work. Handles all three provisioning_mode branches (provision, existing-owned, existing-shared) with infrastructure readiness, CI/CD-triggered deploy, verification, amendment gate, and optional release tagging. No dev-VM sweep, no deploy-rsync skills — the Nth deploy on infrastructure already proven. Also use when the user says "deploy this", "push to production", "ship this feature", or asks to deploy to a project that already has CI/CD workflows from its initial greenfield run.
Use before any local deployment or credential operation that requires secrets from a local HashiCorp Vault instance. Verifies Vault is running, unsealed, and authenticated on 127.0.0.1:8200. Also runs an interactive credential loader script that prompts for all project secrets and writes them to Vault. Use for first-time setup, adding a new project's credentials, or when a Vault path is missing. Also use when setting up Vault for the first time on a new machine, or when a caller cannot reach Vault and needs troubleshooting guidance.
Use when provisioning cloud infrastructure for a project's target application on Cyso Cloud (OpenStack) using Terraform. Covers application credential sourcing, Object Storage backend initialization, SSH keypair variable configuration, `terraform init/plan/apply`, and fixed IP retrieval. Produces a running Cyso VM accessible via SSH, with Cinder data volume attached and security group configured on ports 22/80/443. Also use when re-running Terraform after infrastructure changes or when an operator invokes `/provision.cyso`.
Use when configuring a GitHub repository with the secrets and variables required for CI/CD workflows. Covers Actions secrets (VM_SSH_KEY, Terraform backend credentials, cloud provider credentials, GH_TERRAFORM_TOKEN) and Actions variables (VM_HOST, DEMO_MODE). Uses the `gh` CLI. Prevents GITHUB_TOKEN reserved-name collisions and SSH key trailing-newline drift. Also use when setting up a new deployment repository, rotating CI/CD credentials, or verifying that all required repository secrets and variables are present.
Use when orchestrating the complete SDLC for a new project — from raw project assets to a running production application accessible via browser. Covers requirements extraction, specification (business, functional, stack, infrastructure, coverage), task creation, development, IaC generation + dev environment sweep (local provisioning + deploy + verify), CI/CD production deployment, optional domain/TLS, validation, and tagged release. Re-entrant: use the pre-condition checklist to resume at any phase. Also use when the user says "take this from zero to prod", "run the full smaqit pipeline", "deploy a new project end-to-end", or when starting implementation on a freshly initialized repository.
| name | smaqit.create-agent |
| description | Creates a new agent for this project. Use when the user asks to create, define, or build a new custom agent. |
| metadata | {"version":"2.0.0"} |
Ask the user for the agent name in a single message (lowercase, hyphens allowed, e.g., my-reviewer). The description will be inferred from the name and scanned context.
Before writing anything, read:
.github/agents/ — for naming conventions and tool patterns already used in this projectAlso extract any relevant detail the user has already provided in the conversation (tools, scope constraints, directives).
Using the name and scanned context, infer a complete agent specification. Do not ask further questions.
Write the inferred specification to .smaqit/definitions/agents/[name].md. Create the directory if it does not exist.
For any field where the correct value is genuinely ambiguous, suffix the value with [?] and add a brief inline note explaining the uncertainty.
The definition file must cover:
Invoke smaqit.L2 as a subagent with:
"Compile the agent definition at
.smaqit/definitions/agents/[name].md. Write the compiled agent to.github/agents/[name].agent.md. After compilation, list any fields annotated with[?]and suggest a resolution for each."
After L2 completes, report to the user:
[?]-annotated items and L2's suggested resolutions.smaqit/definitions/agents/[name].md and re-invoke /smaqit.create-agent, or switch to smaqit.L2 directly.smaqit/definitions/agents/[name].md — inferred specification (scaffolding).github/agents/[name].agent.md — compiled agent file (source of truth)Does not create skills, framework files, templates, or Level agents. Does not modify existing agents.
.smaqit/definitions/agents/[name].mdsmaqit.L2 invoked and compilation completed.github/agents/[name].agent.md| Situation | Action |
|---|---|
| Name not provided | Request before proceeding |
.smaqit/templates/ not present | Inform the user that ADK templates are required — run smaqit-adk lite in this repository first |
| Output artifact already exists | Report the conflict; do not overwrite without user confirmation |
| L2 invocation fails | Report the error and include the path to the definition file so the user can inspect or correct it |