create-agent

name	create-agent
description	Scaffold a new Trinity-compatible Claude Code agent from scratch on any topic. Creates directory, CLAUDE.md, skills, and Trinity files — ready for development.
argument-hint	[topic or purpose]
disable-model-invocation	false
user-invocable	true
allowed-tools	Read, Write, Edit, Bash, Glob, Grep, AskUserQuestion, Skill, mcp__trinity__list_agents
metadata	{"version":"1.11","created":"2026-04-01T00:00:00.000Z","updated":"2026-08-18T00:00:00.000Z","author":"Ability.ai","changelog":["1.11: Platform-truth refresh (Trinity v0.9.0, tag 93d7ce7c) — .mcp.json.template rule added: an http/sse server url must resolve to a public address (loopback/private/link-local/CGNAT 100.64/10 = Tailscale refused with 400, no override — trinity-enterprise#394); the report guard in the generated CLAUDE.md also swallows the `requires an agent-scoped API key` refusal a user/admin-key session gets (mcp-server reports.ts)","1.10: template.yaml scaffold now declares `plugins:` (trinity#1704 / ent#411) — marketplaces + installed (agent-dev@abilityai, trinity@abilityai) — so the DEPLOYED agent gets its plugins headlessly on every container boot instead of depending on a human running /plugin install; the local install step stays (that is your own session), the declaration is what makes it portable","1.9: Generated CLAUDE.md Guidelines gain the playbook-call rule — the agent packages procedures as playbooks and exchanges work with other agents only via one-line `/playbook [args]` calls, never prose delegation (fleet convention protocols/playbook-call.md, operator direction 2026-08-16)","1.8: Platform-truth refresh (Trinity dev 88a4e2f7) — report payload cap corrected 256 KB → 5 MiB (object only), display_hint gains `json` and now drives the customer-facing Workspace Reports tab, and list_reports/get_report are taught as read-before-write. template.yaml scaffold gains credentials: + credential_setup: (ent#128/#127; gate T-015). schedules: block documents the ent#89 contract — materialized at creation, max 20, deduped by name, armed only by a literal YAML true, never re-applied on recreate, and gated again by agent autonomy (OFF on new agents); dropped the non-schema id: key and moved timezone off America/New_York to UTC (#1795, and legacy IANA aliases now 500, #1823). .gitignore gains .claude/settings.json + .trinity/* (trinity#2036/#1936) Its .mcp.json.template no longer ships a hand-written `trinity` entry (the platform injects and overwrites its own) and TRINITY_URL/TRINITY_API_KEY are gone from .env.example (read only by the retired CLI); report_type documents the ^[a-z0-9_]+(\\.[a-z0-9_]+)+$ rule — a hyphenated agent name 422s.","1.7: Repository-first deployment — the GitHub-repo step is framed as the deploy path (Trinity clones the repo and tracks the branch; skipping means an upload-only deploy with no reproducible source), and the deploy offer now states what /trinity:onboard actually does: create_agent(template: github:owner/repo@branch) when a remote exists — schedules materialized at creation, updates via git push + git_pull — falling back to a local-file deploy that offers promotion onto the repo path","1.6: Generated CLAUDE.md gains a Request Dispatch section — an SOP table routing incoming requests (user, other agents, operator queue) to skills; task requests with no matching skill are handled if safe and flagged as playbook gaps (told to the user interactively, filed as a playbook-gap-<slug> operator-queue item when headless on Trinity) with a pointer to /agent-dev:create-playbook","1.5: Trinity-connected deploy is the default next action — new Step 14 offers deploying the freshly created agent from its repository via /trinity:onboard when Trinity MCP is connected, gated by explicit AskUserQuestion confirmation; skipped silently when not connected","1.4: Generated agents publish structured reports via the mcp__trinity__report tool — CLAUDE.md gains a Reporting-to-Trinity section, result-producing skills get a guarded report step, and /update-dashboard also emits a kpi_snapshot report (history alongside the live snapshot)","1.3: Scaffold ships README.md + ARCHITECTURE.md + TARGET-ARCHITECTURE.md (current→target development model) and a /reconcile-docs default skill that keeps them coherent with CLAUDE.md, skills, and subagents","1.2: Wizards emit a template.yaml schedules: block for declarative Trinity scheduling","1.1: Removed Trinity CLI references — deployment guidance is now MCP/onboard-based","1.0: Backfilled the /agent-dev:add-git-sync follow-up prompt into the scaffold"]}

Create Agent

ℹ️ First, set expectations: before anything else, print one short line with this skill's version and its most recent change — the top entry of metadata.changelog above — e.g. create-agent vX.Y — recent: <summary>. Then proceed.

Scaffold a complete Claude Code agent from scratch. The agent will be Trinity-compatible and ready for development with playbook-based skill creation.

---

STEP 1: Gather Agent Requirements

If the user provided a topic as an argument, use it as context. Otherwise, ask.

Use AskUserQuestion to gather the following (can be a single open-ended question if user already provided detail, or multiple focused questions):

1a. Agent Purpose

Question: "What should this agent do? Describe its purpose, who it serves, and what problems it solves."

Header: "Agent Purpose"

Get enough detail to write a meaningful CLAUDE.md. Push for specifics — not "a coding agent" but "an agent that reviews Python PRs for a data engineering team and checks for SQL injection, missing tests, and schema migration issues."

1b. Agent Name

Question: "What should the agent be called?"

Header: "Agent Name"

• Suggest a name based on the purpose (lowercase-with-hyphens, short, memorable)
• Let the user override
• This becomes the directory name and the agent identity

1c. Destination

Question: "Where should I create the agent?"

Header: "Location"

Options:

1. ~/[agent-name] — Home directory (recommended)
2. ./[agent-name] — Current directory
3. Custom path — Let me specify

Expand ~ to actual home directory:

echo "$HOME"

1d. Initial Skills

Question: "What should this agent be able to do from day one? List 2-4 key capabilities."

Header: "Starting Skills"

Examples based on the purpose — if it's a content agent, suggest: "generate blog posts, review drafts, manage editorial calendar." If it's an ops agent: "check service health, deploy updates, investigate incidents."

These will become the agent's first skills.

1e. Starter Plugin Selection

Present the available plugins from the Ability.ai marketplace that are relevant to this agent's purpose. Let the user choose which to include in the agent's setup instructions.