agenticx-agent-builder

name	agenticx-agent-builder
description	Guide for creating persistent Near desktop digital avatars (数字分身) via natural-language interview and the create_avatar tool. Use when the user wants to create an avatar, digital twin, specialist agent persona, or add someone to the avatar sidebar.
metadata	{"author":"AgenticX","version":"0.5.0"}

Near Digital Avatar Builder

Help the user create a persistent desktop digital avatar (数字分身) that appears in the Near sidebar avatar list.

Hard rules

1. Must call create_avatar after clarifying and distilling the config. That tool writes the avatar into the registry and makes it appear in the sidebar.
2. Do NOT only write a Markdown file to Desktop / workspace and claim the avatar is created.
3. Do NOT tell the user there is "no create-avatar API" — create_avatar is the API.
4. Do NOT ask the user to open「+ 新建」and fill the manual form unless they explicitly want the GUI.
5. Do NOT invent Python SDK / AgentExecutor / agx agent create workflows for this task — those are code-layer agents, not Near desktop avatars.

Interview flow (2–3 rounds, not a long questionnaire)

Ask only what is still missing. Prefer short choices (clarification / show_widget) over open essays.

Round 1 — Positioning

• What should this avatar own? (domain / job)
• Who is it for? (usually the current user)

Round 2 — Scope & style

• Coverage (markets, stacks, products, languages…)
• Analysis / work style (trading vs research, concise vs thorough, etc.)

Round 3 — Boundaries (only if needed)

• Risk preference / what it must never do
• Preferred output structure
• Optional default model

Stop early when you already have enough to write a solid name / role / system_prompt.

Distill into these fields

Field	Meaning	Example
name	Short character-like display name	飞廉
role	One-line positioning	进取型全市场交易分析师
system_prompt	Full behavior contract	See template below
default_provider / default_model	Optional; omit to inherit global default

system_prompt should include

• Identity & coverage
• Analysis / work priorities
• Risk preference & hard boundaries
• Data discipline (no fabricated numbers / facts)
• Output style & fixed structure when useful
• Disclaimer when the domain needs one (finance, medical, legal, etc.)

Landing step (required)

Call:

create_avatar(
  name=...,
  role=...,
  system_prompt=...,
  default_provider=optional,
  default_model=optional
)

Then confirm to the user:

• The avatar is in the sidebar list
• They can open it or @ it in group chat
• Optionally offer follow-ups: schedule automation, tweak prompt, set default model

If create_avatar returns avatar_exists

Do not create a duplicate. Tell the user it already exists and offer:

• delegate_to_avatar(avatar_id=..., task=...) for work
• Or update via settings / a later update tool if they want changes

Anti-patterns

• Saying "半搞定" after only drafting a prompt file
• Dumping a long YAML/Markdown archive without calling create_avatar
• Asking the user to copy-paste into the create-avatar form
• Confusing temporary spawn_subagent workers with persistent avatars
• Using spawn_subagent with the same name as a registered avatar

Example success path

User: "帮我创建一个数字分身专门帮我分析金融市场"

1. Clarify markets / style / risk preference (2–3 choices)
2. Distill name=飞廉, role=进取型全市场交易分析师, full system_prompt
3. Call create_avatar(...)
4. Reply: 飞廉已加入分身列表，可直接点开或 @ 它