setup

/hypo-workflow:setup

Output Language Rules

📌 输出语言规则： 读取 config.yaml → output.language

• zh-CN / zh：所有用户可见的输出使用中文（PROGRESS、报告、状态提示、错误消息、交互提问）
• en：使用英文
• auto：跟随用户对话语言 内部日志（log.yaml、state.yaml）始终英文。

Use this skill as the plugin-level setup wizard. It configures Hypo-Workflow itself, not a project-local .pipeline/ workspace.

Configuration Target

Always use the global file:

• directory: ~/.hypo-workflow/
• config file: ~/.hypo-workflow/config.yaml

Do not write .pipeline/config.yaml from setup. Project-local configuration belongs to /hypo-workflow:init or /hypo-workflow:plan-generate.

Config Priority

When another skill needs runtime defaults, resolve values in this order:

1. project config: .pipeline/config.yaml
2. global config: ~/.hypo-workflow/config.yaml
3. built-in defaults

Important mappings:

• project execution.mode > global execution.default_mode > self
• project execution.subagent_tool > global subagent.provider > auto
• project plan.mode > global plan.default_mode > interactive
• project plan.interaction_depth > global plan.interaction_depth > medium
• project dashboard.enabled > global dashboard.enabled > false
• project dashboard.port > global dashboard.port > 7700
• project output.language > global output.language > zh-CN
• project output.timezone > global output.timezone > Asia/Shanghai
• project watchdog.enabled > global watchdog.enabled > false
• project rules from .pipeline/rules.yaml > project rules.* > global rules.* > extends: recommended

Execution Flow

1. Create the config directory if needed:
   • mkdir -p ~/.hypo-workflow
2. Check for an existing global config:
   • if ~/.hypo-workflow/config.yaml does not exist, start the first-run wizard
   • if it exists, read it, summarize the current values, and ask whether the user wants to modify it
3. Detect the current agent platform:
   • Claude Code when Claude-specific markers or environment are visible
   • Codex when Codex-specific markers or environment are visible
   • otherwise ask the user to choose claude-code or codex
4. Ask for execution mode:
   • self is recommended for first-time users
   • subagent enables delegation to another agent runtime
5. If subagent mode is selected, ask for:
   • provider: codex or claude
   • model
   • Codex base_url only when a custom API endpoint is used
   • optional connection test command when the tool is available
6. Ask for Dashboard defaults:
   • enabled by default: true or false
   • port, default 7700
7. Ask for Plan mode:
   • interactive pauses for confirmation during planning
   • auto runs planning phases without routine pauses
8. Ask for output defaults:
   • language, default zh-CN
   • timezone, default Asia/Shanghai
9. Ask whether to enable watchdog defaults:
   • default false
   • interval and heartbeat timeout default 300
10. Ask for Rules defaults:

• preset: recommended, strict, or minimal
• default recommended

11. Write ~/.hypo-workflow/config.yaml.
12. Print a concise configuration summary and remind the user that project config can override these values.

First-Run Questions

Use short, concrete prompts:

• "Current platform appears to be <detected>. Use this platform?"
• "Default execution mode: self or subagent?"
• "Subagent provider: codex or claude?"
• "Subagent model?"
• "Custom Codex base URL, or leave empty?"
• "Enable Dashboard by default?"
• "Dashboard port?"
• "Plan mode: interactive or auto?"
• "Interaction depth: low, medium, or high?"
• "Output language?"
• "Output timezone?"
• "Enable Auto Resume watchdog by default?"
• "History import split method?"
• "Rules preset: recommended, strict, or minimal?"

Default Values

• agent.platform=claude-code when running in Claude Code, otherwise codex
• agent.model should use the current session model when visible
• execution.default_mode=self
• subagent.provider=codex
• subagent.codex.model=gpt-5.4
• subagent.claude.model=claude-sonnet-4-20250514
• dashboard.enabled=true
• dashboard.port=7700
• plan.default_mode=interactive
• plan.interaction_depth=medium
• plan.interactive.min_rounds=3
• plan.interactive.require_explicit_confirm=true
• output.language=zh-CN
• output.timezone=Asia/Shanghai
• watchdog.enabled=false
• watchdog.interval=300
• watchdog.heartbeat_timeout=300
• watchdog.max_retries=5
• watchdog.max_consecutive_milestones=10
• watchdog.notify=true
• history_import.split_method=auto
• history_import.time_gap_threshold=24h
• history_import.max_milestones=20
• compact.auto=true
• compact.progress_recent=15
• compact.state_history_full=1
• compact.log_recent=20
• compact.reports_summary_lines=3
• showcase.language=auto
• showcase.poster.api_key_env=OPENAI_API_KEY
• showcase.poster.size=1024x1536
• showcase.poster.quality=high
• showcase.poster.style=auto
• rules.extends=recommended
• rules.rules={}
• version=8.4.0

Config Shape

# Hypo-Workflow global config
# Generated by /hypo-workflow:setup

agent:
  platform: claude-code
  model: claude-sonnet-4-20250514

execution:
  default_mode: self

subagent:
  provider: codex
  codex:
    model: gpt-5.4
    base_url: https://api.vsplab.cn
  claude:
    model: claude-sonnet-4-20250514

dashboard:
  enabled: true
  port: 7700

plan:
  default_mode: interactive
  interaction_depth: medium
  interactive:
    min_rounds: 3
    require_explicit_confirm: true

output:
  language: zh-CN
  timezone: Asia/Shanghai

watchdog:
  enabled: false
  interval: 300
  heartbeat_timeout: 300
  max_retries: 5
  max_consecutive_milestones: 10
  notify: true

history_import:
  split_method: auto
  time_gap_threshold: 24h
  max_milestones: 20
  keyword_patterns:
    - 'feat\(M(\d+)\):'
    - 'M(\d+)-'
    - 'milestone-(\d+)'

compact:
  auto: true
  progress_recent: 15
  state_history_full: 1
  log_recent: 20
  reports_summary_lines: 3

showcase:
  language: auto
  poster:
    api_key_env: OPENAI_API_KEY
    size: "1024x1536"
    quality: high
    style: auto

rules:
  extends: recommended
  rules: {}

version: "8.4.0"
created: "2026-04-26T14:00:00+08:00"
updated: "2026-04-26T14:00:00+08:00"

Omit subagent.codex.base_url when the default OpenAI endpoint is used.

Connection Checks

Only run checks that match installed tools:

• Codex provider:
  • codex --version
  • if custom API variables are needed, tell the user to set OPENAI_BASE_URL and OPENAI_API_KEY
• Claude provider:
  • claude --version

If a check fails, keep the config but mark the provider as unverified in the summary. Do not block setup unless the user asks for a strict check.

Platform Notes

Claude Code users:

• invoke commands as /hypo-workflow:<command>
• configure Codex as a subagent by installing @openai/codex, setting OPENAI_API_KEY, and choosing subagent.provider=codex
• use /hypo-workflow:dashboard for the WebUI

Codex users:

• invoke the compatibility commands as /hw:*
• configure Claude as a subagent by installing @anthropic-ai/claude-code and choosing subagent.provider=claude
• keep project-specific overrides in .pipeline/config.yaml

Mixed mode:

• keep execution.mode=self for the main orchestrator
• delegate individual steps with step_overrides.<step>.executor=subagent
• set step_overrides.<step>.subagent=codex or claude

Existing Config Behavior

When a config exists:

1. show current values for platform, default mode, subagent provider/model, dashboard, and plan mode
2. ask whether to edit
3. preserve unspecified existing values

Config TUI Editing

Interactive config editing must keep global defaults and project config as separate targets. Stage edits first, show a diff, validate supported schema fields, and require confirmation before writing.

Do not write .pipeline/state.yaml, .pipeline/cycle.yaml, or .pipeline/rules.yaml from the setup/config TUI. If adapter-affecting values change, tell the user to run /hw:sync --light. 4. update updated 5. keep the original created timestamp

Reference Files

• references/config-spec.md - global/project config priority and field mapping
• config.schema.yaml - project schema plus global config schema definition
• SKILL.md - full system reference if broader context is needed