bootstrap-existing-agent-with-prefactor-cli（bootstrap existing Agent with prefactor CLI 命令行）

Bootstrap Existing Agent With Prefactor CLI

Set up Prefactor resources for an already-working agent before instrumentation code changes.

Core principle: provision first, instrument second.

Coding Assistant Usage

Apply this skill first when the user asks to:

• "set up Prefactor for this existing agent"
• "create Prefactor environment/agent/instance"
• "use CLI to bootstrap Prefactor"
• "prepare IDs and env vars before instrumentation"

After this skill completes:

1. If provider is supported, continue with skills/instrument-existing-agent-with-prefactor-sdk/SKILL.md.
2. If provider is unsupported, continue with skills/create-provider-package-with-core/SKILL.md.
3. Return a copy/paste block with exported env vars and the selected package.

Inputs You Need

• Prefactor API token (for CLI profile)
• Base URL (optional, defaults to Prefactor cloud)
• Account ID
• Target provider/framework (langchain, ai, openclaw, or custom)
• Human-readable names for environment and agent
• Working directory to store config (recommended: repo root)

CLI Workflow

Before running CLI commands, choose package first, then install required Prefactor package(s).

• Use whichever package manager the project already uses (bun, npm, pnpm, or yarn).
• Install @prefactor/cli for bootstrap commands.

prefactor command requirement:

• The prefactor command comes from the npm package @prefactor/cli.
• If the command is not globally available, run it via the package manager launcher (bunx @prefactor/cli, npx @prefactor/cli, pnpm dlx @prefactor/cli, or yarn dlx @prefactor/cli).
• Use prefactor help or prefactor <group> help for command details.

Examples:

# bun
bun add @prefactor/cli

# npm
npm install @prefactor/cli

# pnpm
pnpm add @prefactor/cli

# yarn
yarn add @prefactor/cli

Run these in order:

prefactor profiles add default [base-url] --api-token <api-token>
prefactor accounts list
prefactor environments create --name <env-name> --account_id <account-id>
prefactor agents create --name <agent-name> --environment_id <environment-id>
prefactor agent_instances register \
  --agent_id <agent-id> \
  --agent_version_external_identifier <agent-version-id> \
  --agent_version_name <agent-version-name> \
  --agent_schema_version_external_identifier <schema-version-id> \
  --update_current_version

Profile notes:

• <profile-name> is any key like default, staging, or prod.
• Select profile with --profile <name>.
• When using launchers, prefix commands consistently (for example npx @prefactor/cli profiles add...).

Config resolution notes:

• CLI config resolution order is:

1. ./prefactor.json 2. ~/.prefactor/prefactor.json 3. if none exists, profile creation writes ./prefactor.json

• Global CLI install does not make config global; command working directory still controls which config file is used.

Collect and persist these IDs from command output:

• environment_id
• agent_id
• agent_instance_id

Package Selection

Choose package by provider:

• LangChain -> @prefactor/langchain
• AI SDK -> @prefactor/ai
• OpenClaw -> @prefactor/openclaw
• Custom/unsupported provider -> use skills/create-provider-package-with-core/SKILL.md

When handing off to SDK instrumentation, import helpers from that selected package directly, for example:

import { init, withSpan, shutdown } from '@prefactor/ai';
// or '@prefactor/langchain'

Do not mix adapter init with withSpan/shutdown from @prefactor/core unless an explicit tracer is passed. This guidance targets adapter-style integrations (@prefactor/ai, @prefactor/langchain) and does not change @prefactor/openclaw plugin runtime behavior.

If you have identified and selected an existing package, use skills/instrument-existing-agent-with-prefactor-sdk/SKILL.md

Runtime Environment Variables

Produce this output for the user after setup:

export PREFACTOR_API_URL="<api-url>"
export PREFACTOR_API_TOKEN="<api-token>"
export PREFACTOR_AGENT_ID="<agent-id>"

Use the created agent_id for PREFACTOR_AGENT_ID.

Verification

• Confirm CLI commands succeeded without HTTP/auth errors.
• Confirm IDs were returned and captured.
• Confirm package selection matches provider.
• Confirm env vars match created resources.
• Confirm prefactor.json is ignored by git (git check-ignore prefactor.json, git status --short).

Common Mistakes

• Instrumenting code before creating Prefactor resources.
• Using account ID where environment ID is required.
• Forgetting to propagate created agent_id to PREFACTOR_AGENT_ID.
• Picking @prefactor/core directly when a built-in adapter exists.
• Running commands from the wrong directory and reading/writing the wrong prefactor.json.
• Committing prefactor.json (contains API tokens).