devflow-setup

DevFlow Setup

Say exactly: "SKILL INVOKED: setup"

Critical Rules

• Check first, ask later — detect MCP access before asking configuration questions
• Never guess or fabricate MCP parameters (cloudId, project IDs, emails, etc.)
• Always confirm detected settings with user before storing
• Each reference file is self-contained — follow it completely

---

Step 0: Check Existing Configuration

Check for an existing config file:

1. Check ~/.claude/plugins/config/devflow/config.md

If config exists, tell the user:

"You have an existing DevFlow configuration. Would you like to:"

1. Reconfigure from scratch
2. Keep current settings
3. View current settings

If "Keep" → end. If "View" → display config, then ask if they want to reconfigure.

---

Step 1: Issue Tracking

Ask:

"Do you want to use an issue tracker with DevFlow?"

1. Yes - Jira (Atlassian MCP)
2. Yes - GitLab Issues (GitLab MCP)
3. Yes - GitHub Issues (gh CLI)
4. No - Skip issue tracking

Route based on selection:

• Jira → follow references/issues-jira.md
• GitLab → follow references/issues-gitlab.md
• GitHub → follow references/issues-github.md
• Skip → set ISSUES_BACKEND=none, ISSUES_ENABLED=false

Store results: ISSUES_BACKEND, ISSUES_ENABLED, ATLASSIAN_VERIFIED (bool), CLOUD_ID (if Jira)

---

Step 2: Documentation System

Ask:

"Do you want documentation integration?"

1. Yes - Confluence (Atlassian)
2. Yes - Google Drive
3. No - Skip documentation

Route based on selection:

• Confluence:
  • If ATLASSIAN_VERIFIED=true from Step 1: "Atlassian MCP already connected. Confluence is available." Ask for optional default_space, then continue.
  • If NOT verified → follow references/docs-confluence.md
• Google Drive → follow references/docs-google-drive.md
• Skip → set DOCS_BACKEND=none, DOCS_ENABLED=false

---

Step 3: RAG Memory

Ask:

"Do you want RAG Memory for AI-retrievable knowledge?"

RAG Memory stores knowledge that AI can search and retrieve — useful for technical documentation, project notes, and quick references.

1. Yes - Enable RAG Memory
2. No - Skip RAG Memory

Route based on selection:

• Yes → follow references/rag-memory.md
• Skip → set RAG_ENABLED=false

If both docs AND RAG Memory enabled, ask routing preference:

"You have both a docs backend and RAG Memory. How should content be routed?"

1. Let me decide each time (default)
2. Quick notes → RAG, formal docs → [Confluence/Google Drive]
3. Everything to RAG Memory

---

Step 4: Code Intelligence (LSP Plugins)

Ask:

"Do you want to enable Code Intelligence with LSP plugins?"

LSP plugins give Claude real-time type checking, jump-to-definition, and immediate error detection for your code.

1. Yes - Set up LSP plugins for languages I use
2. No - Skip code intelligence

Route based on selection:

• Yes → follow references/lsp-plugins.md
• No → set LSP_ENABLED=false

Store results: LSP_ENABLED, installed_plugins[], pending_plugins[]

---

Step 5: VCS Configuration

VCS is determined by issue tracker choice:

• GitLab Issues selected → VCS = GitLab (MRs)
• GitHub Issues selected → VCS = GitHub (PRs)
• Jira or none → auto-detect from git remote:

git remote get-url origin 2>/dev/null

• Contains github.com → GitHub (PRs)
• Contains gitlab → GitLab (MRs)
• Other → default GitHub

Display:

"VCS Backend: [GitHub PRs / GitLab MRs]"

---

Step 6: Summary & Confirmation

Display a summary table of all settings:

DevFlow Configuration Summary
=============================

Issue Tracking:
  Backend: [jira/gitlab/github/none]
  Status: [enabled/disabled]

Documentation:
  Backend: [confluence/google-drive/none]
  Status: [enabled/disabled]
  Organization: [manual/default-folder/ask-each-time]

Knowledge Base (RAG Memory):
  Status: [enabled/disabled]
  Routing: [decide-each-time/quick-to-rag/all-to-rag]

Code Intelligence (LSP):
  Status: [enabled/disabled]
  Installed Plugins: [list of installed LSP plugins]

Version Control:
  Backend: [github/gitlab]
  Auto-detected: [yes/no]

Ask:

"Save this configuration?"

1. Yes - Save configuration
2. No - Cancel setup

---

Step 7: Write Configuration

First, create the config directory if it doesn't exist:

mkdir -p ~/.claude/plugins/config/devflow

Then write to ~/.claude/plugins/config/devflow/config.md:

# DevFlow Backend Configuration
# Generated by /devflow-setup
# Regenerate anytime with: /devflow-setup

## Issue Tracking
backend: [jira/gitlab/github/none]
enabled: [true/false]
# For Jira:
cloudId: [if jira]
# For GitLab:
default_project: [if gitlab, optional]
# For GitHub:
# Uses current repo context via gh CLI

## Documentation
backend: [confluence/google-drive/none]
enabled: [true/false]
organization: [manual/default-folder/ask-each-time]
# For Google Drive with default-folder:
default_folder_id: [folder id]
# For Confluence:
default_space: [space key, optional]

## Knowledge Base (RAG Memory)
enabled: [true/false]
routing: [decide-each-time/quick-to-rag/all-to-rag]

## Code Intelligence (LSP)
enabled: [true/false]
installed_plugins: [list]
pending_plugins: [list with binaries needed]

## Version Control
backend: [github/gitlab]
# Auto-detected from git remote

---

Complete

Say: "DevFlow configured! Your backends are ready."

Discover Available Skills

1. Find the installed plugin location:

   # Find plugin cache path
   find ~/.claude/plugins/cache/claude-code-primitives -name "devflow" -type d 2>/dev/null | head -1
   

   Or if in source repo: plugins/devflow

2. List all skills in the plugin's skills/ directory:

   ls "$PLUGIN_PATH/skills/"
   

3. For each skill, follow the symlink and read SKILL.md to extract:

   • name from frontmatter
   • description from frontmatter

4. Group skills by category prefix (build, pm, docs, rag-memory, devops) and present:

## Available Skills

### Build ([count] skills)
| Skill | Description |
|-------|-------------|
| `/devflow:build:fetch-issue` | description |

### PM ([count] skills)
...

5. End with:

Reconfigure anytime: /devflow-setup

Full guide: /devflow:plugin-overview to see all skills, hooks, and usage

If you enabled RAG Memory: Run /devflow:rag-memory:setup-collections to create recommended collections

---

Companion Plugins

Recommend these after setup is complete. Install via /plugin install <name>@claude-plugins-official.

Plugin	Required?	Purpose
context7	Required for tech research	Provides up-to-date library docs for the tech-researcher agent used in architecture design. Without it, tech research falls back to training data only.
playground	Recommended	Generates interactive HTML code maps and diff viewers. Used optionally in fetch-issue (explore codebase before planning), document-legacy-codebase (visual code map), and review-pr (interactive diff visualization).
claude-md-management	Recommended	Audits and improves CLAUDE.md files over time. Complements DevFlow's generate-claude-md skill — DevFlow creates the convention layer, claude-md-management keeps it current. Run /revise-claude-md at the end of any session to capture what was learned.

Install all three:

/plugin install context7@claude-plugins-official
/plugin install playground@claude-plugins-official
/plugin install claude-md-management@claude-plugins-official