kkclaude

kkclaude

Personal Claude Code plugin for Jira ticket drafting and engineering proposal writing.

Structure

kkclaude/
├── .claude-plugin/          # Plugin manifest
├── agents/                  # Autonomous agent definitions
├── skills/                  # Skill slash commands
├── hooks/                   # Lifecycle automation scripts
│   ├── hooks.json
│   └── scripts/
│       └── lib/
│           └── mcp.js       # mcporter wrapper for hook scripts
├── config/
│   └── mcporter.json        # mcporter config (imports .mcp.json servers)
├── package.json
└── CLAUDE.md

Skills (Slash Commands)

Command	Description
/kkclaude:jira-ticket	Draft a Jira ticket with Overview, Context, AC
/kkclaude:engineering-proposal	Draft an engineering proposal, RFC, or architecture ADR

Agents

Agent	Model	Purpose
kkclaude:ticket-create	sonnet	Draft Jira tickets in Jira wiki markup
kkclaude:proposal-writer	sonnet	Draft engineering proposals and architecture decision docs

Hooks

Event	Script	Purpose
SessionStart	session-start.js	Log session info + fetch in-progress Jira tickets
UserPromptSubmit	user-prompt-submit.js	Inject git context into every prompt
PreToolUse[Bash]	bash-safety-check.js	Block dangerous shell commands
PostToolUse[Edit|Write]	auto-format.js	Auto-format edited files on save

MCP Integration (mcporter)

Hook scripts call MCP tools (Atlassian/Jira, Serena) via mcporter — a lightweight JS runtime that auto-discovers servers from .mcp.json.

config/mcporter.json uses "imports": ["claude-code"] so .mcp.json is the single source of truth — no config duplication.

Available MCP Servers

Server	Type	Purpose
atlassian	HTTP/OAuth	Jira, Confluence, and Compass
serena	stdio	Semantic code navigation and editing (IDE-like)

Both are defined in .mcp.json and picked up automatically by mcporter via "imports": ["claude-code"].

Installation

mcporter is included in package.json. Install with:

npm install

Or install globally for CLI access:

brew install steipete/tap/mcporter
# or
pnpm add -g mcporter

Serena Setup

Serena requires uv to be installed (Python package manager used to run uvx):

brew install uv

Ensure ~/.mcporter/mcporter.json contains the following entry under mcpServers:

{
  "mcpServers": {
    "serena": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/oraios/serena",
        "serena",
        "start-mcp-server",
        "--context",
        "claude-code",
        "--project-from-cwd",
        "--open-web-dashboard",
        "False"
      ],
      "description": "Serena semantic code toolkit for IDE-like code navigation and editing. Optimized for large repos (30k+ files). Use context 'claude-code' to avoid tool duplication with Claude Code built-ins."
    }
  }
}

Serena launches on-demand via uvx — no further installation needed. Verify it works:

npx mcporter list serena

Test a Serena tool call:

npx mcporter call serena.get_symbols_overview relative_path:'src/'

Atlassian OAuth Setup

The atlassian MCP server uses OAuth. Authenticate once before using Jira-enriched hooks:

npx mcporter auth atlassian

This opens a browser login and caches the token under ~/.mcporter/atlassian/.

Verify it works and retrieve your Cloud ID (required for every Atlassian tool call):

npx mcporter call atlassian.getAccessibleAtlassianResources

Copy the id field from the response. Set it as an environment variable so hooks can use it:

export ATLASSIAN_CLOUD_ID=<id-from-above>
# Add to ~/.zshrc or ~/.zshenv to persist across sessions

Troubleshooting

Session-start Jira fetch silently skipped

The SessionStart hook attempts to fetch in-progress Jira tickets with a 3-second timeout. It is silently skipped when: