Skip to content

plugins.md

Latest commit

 

History

History
176 lines (121 loc) · 6.92 KB

plugins.md

File metadata and controls

176 lines (121 loc) · 6.92 KB

Plugin System

CodeForge includes a local plugin marketplace that provides specialized tools and hooks for Claude Code.

Architecture

plugins/devs-marketplace/
├── .claude-plugin/
│   └── marketplace.json       # Marketplace manifest (lists all plugins)
└── plugins/
    ├── codeforge-lsp/         # LSP language servers
    ├── ticket-workflow/       # EARS ticket workflow
    ├── notify-hook/           # Desktop notifications
    ├── dangerous-command-blocker/  # Safety: block destructive commands
    ├── protected-files-guard/ # Safety: protect sensitive files
    ├── auto-code-quality/     # Batch formatter + linter + advisory test runner
    ├── agent-system/          # 17 custom agents + redirection hooks
    ├── skill-engine/          # 22 coding skills + auto-suggestion
    ├── spec-workflow/         # 8 spec lifecycle skills + spec-reminder
    ├── session-context/       # Git state, TODO harvesting, commit reminders
    └── workspace-scope-guard/ # Workspace scope enforcement

Each plugin has a .claude-plugin/plugin.json manifest defining its name, description, and capabilities.

Enabling/Disabling Plugins

Plugins are enabled in config/defaults/settings.json under enabledPlugins:

"enabledPlugins": [
    "auto-code-quality@devs-marketplace",
    "agent-system@devs-marketplace",
    "skill-engine@devs-marketplace",
    "spec-workflow@devs-marketplace",
    "session-context@devs-marketplace"
]

To disable a plugin, remove its entry from the enabledPlugins array.

PLUGIN_BLACKLIST

To skip a plugin during the installation/registration step (without editing settings.json), add it to PLUGIN_BLACKLIST in .env:

PLUGIN_BLACKLIST="ticket-workflow,auto-linter"

This prevents the plugin from being registered on container start but doesn't remove it from enabledPlugins.

Plugin Reference

codeforge-lsp

Purpose: Provides Language Server Protocol servers for code intelligence.

Servers:

  • Pyright — Python type checking and completion (.py, .pyi)
  • TypeScript Language Server — TypeScript/JavaScript (.ts, .tsx, .js, .jsx, .mts, .cts, .mjs, .cjs)
  • gopls — Go language server (.go, .mod, .sum)

Claude Code automatically uses these for hover info, go-to-definition, and find-references.

ticket-workflow

Purpose: EARS-based ticket workflow with GitHub integration.

Provides structured ticket management using EARS (Easy Approach to Requirements Syntax) format. Integrates with GitHub Issues for tracking.

Commands: /ticket:new, /ticket:work, /ticket:review-commit, /ticket:create-pr

notify-hook

Purpose: Desktop notifications when Claude Code finishes responding.

Sends an OSC escape sequence and terminal bell when Claude completes a response, triggering desktop notifications in supported terminals (WezTerm, iTerm2, VS Code with the OSC notifier extension).

Configuration: Enabled via devcontainer.json feature options:

"./features/notify-hook": {
    "enableBell": true,
    "enableOsc": true
}

dangerous-command-blocker

Purpose: Prevents Claude Code from executing destructive bash commands.

Blocked patterns:

  • rm -rf / and variants
  • sudo rm on system directories
  • chmod 777 on sensitive paths
  • git push --force to protected branches
  • dd with output to block devices
  • Other destructive system commands

The blocker runs as a PreToolUse hook on Bash commands. It checks the command against a pattern list and rejects matches.

protected-files-guard

Purpose: Prevents Claude Code from modifying sensitive files.

Protected patterns:

  • .env and .secrets files
  • Lock files (package-lock.json, uv.lock, Cargo.lock, etc.)
  • .git/ directory contents
  • Credential files and SSH keys

Runs as a PreToolUse hook on Write and Edit operations.

auto-code-quality

Purpose: Batch-formats, lints, and runs advisory tests on files edited during a Claude Code session.

Supported formatters:

Language Formatter
Python Ruff
Go gofmt
JavaScript/TypeScript/CSS/JSON/GraphQL/HTML Biome
Shell scripts shfmt
Markdown/YAML/TOML/Dockerfile dprint
Rust rustfmt

Supported linters:

Language Linter
Python Pyright + Ruff
JavaScript/TypeScript/CSS/GraphQL Biome
Shell scripts ShellCheck
Go go vet
Dockerfile hadolint
Rust clippy

How it works: Runs as a Stop hook. When Claude Code stops, it checks which files were edited, detects their language, and runs the appropriate formatter and linter. Also includes an advisory test runner that runs affected tests. Results are informational — they don't block.

agent-system

Purpose: 17 specialized agent definitions with built-in agent redirection.

Components:

  • 17 custom agents — Specialized agent definitions for different task types (architect, explorer, test-writer, refactorer, security-auditor, researcher, doc-writer, etc.)
  • Agent redirection hook — Transparently swaps built-in agent types to custom agents (e.g., Explore → explorer, Plan → architect)
  • CWD injection hook — Injects current working directory into agent prompts
  • Read-only bash enforcement — Prevents read-only agents from executing write operations

For detailed agent documentation, see plugins/devs-marketplace/plugins/agent-system/agents/.

skill-engine

Purpose: 22 domain-specific coding reference skills with auto-suggestion.

Skills: fastapi, svelte5, docker, docker-py, pydantic-ai, sqlite, testing, debugging, security-checklist, refactoring-patterns, git-forensics, performance-profiling, documentation-patterns, migration-patterns, dependency-management, claude-code-headless, claude-agent-sdk, ast-grep-patterns, api-design, skill-building, team, worktree

How it works: Skills are loaded on demand via the Skill tool. A PreToolUse hook auto-suggests relevant skills based on conversation context.

For skill details, see plugins/devs-marketplace/plugins/skill-engine/skills/.

spec-workflow

Purpose: 8 spec lifecycle skills with spec-reminder hook.

Skills: spec-new, spec-refine, spec-build, spec-review, spec-update, spec-check, spec-init, specification-writing

How it works: Provides a structured specification workflow. A Stop hook reminds users to update specs when code was modified but specs weren't.

workspace-scope-guard

Purpose: Enforces workspace scope by blocking writes outside the working directory and warning on out-of-scope reads.

Runs as a PreToolUse hook on Write, Edit, and Read operations. Compares file paths against the current working directory and rejects modifications to files outside the project scope. Read operations outside scope produce a warning but are not blocked. Resolves symlinks and worktree paths correctly via os.path.realpath().