Development Workflow Plugin for Claude Code

A comprehensive Claude Code plugin for feature development, bug fixes, and PR workflows. Enforces best practices through upfront planning, parallelized subagent execution, verification agents, and workflow commands.

Features

Upfront planning - Brainstorming and structured implementation plans before code
Plan execution - Parallel implementation via agent teams (worktrees + mailbox) or subagents (run_in_background), auto-selected
Git worktree support - Isolated workspaces for parallel development
Pre-commit verification - Type checking, security scans, debug code detection
PR review verification - SHA-anchored sticky-comment check + blocker scanning (inline gh ops, no slash command wrapper)
Git guardrails - Blocks direct pushes to main, warns on raw merge commands
Code review agents - Staff-level comprehensive reviews
Context recovery - Restore state after context compaction

Philosophy

This plugin enforces three core principles:

1. Planning Before Implementation

Features start with brainstorming (Socratic exploration of requirements and design), followed by writing plans that break work into bite-sized, verifiable tasks. This catches issues early and provides clear success criteria.

2. Parallel Execution with Mode Selection

Plan execution automatically selects the best execution mode:

Team mode (when agent teams enabled): Multiple implementers in worktrees, dedicated reviewer with mailbox communication, lead in delegate mode
Subagent mode (default): Parallel dispatch via run_in_background, each subagent gets only what you pass it (lightweight context), independent model selection per task

Subagent mode is more efficient for well-scoped tasks — subagents start fresh with just their prompt, avoiding full context loading overhead.

3. Parallelization Through Isolation

Git worktrees + agents enable true parallel development:

Multiple features/experiments in isolated workspaces
Agents work independently without blocking
Easy cleanup (delete worktree, no branch pollution)

Complete Development Workflow

flowchart TD
    subgraph FEATURE ["Feature Development Path"]
        A[New feature request] --> B[Brainstorming skill]
        B --> C["Explore requirements & design"]
        C --> D[Using-git-worktrees skill]
        D --> E[Create isolated workspace]
        E --> F[Writing-plans skill]
        F --> G["Break into tasks with acceptance criteria"]
        G --> H{Agent teams enabled?}
        H -->|Yes| I1["Plan-execution: team mode"]
        I1 --> I2["Parallel implementers + reviewer in worktrees"]
        H -->|No| I3["Plan-execution: subagent mode"]
        I3 --> I4["Parallel dispatch with lightweight context"]
        I2 --> J["Push + create PR"]
        I4 --> J
        J --> K[PR ready]
    end

    subgraph BUG ["Bug Fix Path"]
        L[Bug report] --> M[Systematic-debugging skill]
        M --> N["4-phase root cause analysis"]
        N --> O[Test-driven-development skill]
        O --> P["Red → Green → Refactor"]
        P --> Q[Commit fix]
        Q --> R["Push + create PR"]
        R --> K
    end

    subgraph MERGE ["PR Merge Workflow"]
        K --> S[CI + Reviews]
        S --> T{Approved?}
        T -->|No| U[Address feedback]
        U --> S
        T -->|Yes| V["Verify CI + SHA-anchored sticky comment"]
        V --> W["gh pr checks --watch + match Reviewed commit: SHA to head"]
        W --> X{Blockers?}
        X -->|Yes| U
        X -->|No| Y[Merge & cleanup]
    end

    style FEATURE fill:#e3f2fd
    style BUG fill:#fff3e0
    style MERGE fill:#fce4ec

Feature Development Path

Phase	Skill/Command	Purpose
1. Design	`brainstorming`	Socratic exploration of requirements and design trade-offs
2. Isolate	`using-git-worktrees`	Create isolated workspace (optional, for parallel work)
3. Plan	`writing-plans`	Break feature into bite-sized tasks with acceptance criteria
4. Execute	`plan-execution`	Parallel via agent teams or subagents (auto-selected)
5. PR & Merge	Native `gh` ops	`gh pr checks --watch`, SHA-match latest sticky comment, fix loop until ready, merge

Bug Fix Path

Phase	Skill/Command	Purpose
1. Analyze	`systematic-debugging`	4-phase root cause analysis (gather, hypothesize, verify, fix)
2. Test	`test-driven-development`	Red-green-refactor discipline
3. Commit	N/A	Commit fix with conventional message
4. PR & Merge	Native `gh` ops	Create PR, `gh pr checks --watch`, SHA-match sticky comment, merge

Git Guards

The plugin prevents common mistakes:

Action	On Main	On Feature Branch
`git commit`	Blocked	Allowed
`git push origin main`	Blocked	N/A
`gh pr merge`	Warned (verify CI + SHA match)	Warned (verify CI + SHA match)

Execution Preference

When executing implementation plans (from writing-plans or similar), use dev-workflow:plan-execution. It automatically selects the best mode:

Team mode (agent teams enabled): parallel teammates in worktrees with mailbox communication
Subagent mode (default): parallel dispatch via run_in_background with lightweight context per task

Installation

Option 1: Direct from GitHub (Recommended)

/plugin marketplace add tombakerjr/claude-code-workflows
/plugin install dev-workflow@claude-code-workflows

Then restart Claude Code.

Option 2: From Local Clone

git clone https://github.com/tombakerjr/claude-code-workflows.git
/plugin marketplace add /path/to/claude-code-workflows
/plugin install dev-workflow@tombakerjr-claude-tools

Then restart Claude Code.

Components

Slash Commands

Command	Description
`/context-recovery`	Recover git/PR state after context compaction

Agents

Agent	Description
`staff-code-reviewer`	Comprehensive review: security, correctness, performance, architecture
`quick-reviewer`	Fast combined spec+quality review for simple tasks (≤2 files)
`code-verifier`	Pre-commit: typecheck, security scan, debug code detection
`pr-verifier`	Pre-merge: CI status, comment wait, blocker detection
`implementer`	Task implementation agent (Sonnet default, Haiku for mechanical work)
`spec-reviewer`	Verify implementation matches spec and acceptance criteria
`quality-reviewer`	Fast quality gate: code quality, test coverage, documentation

Skills

Skill	Description
`plan-execution`	Execute implementation plans — agent teams (worktrees + mailbox) or subagents (`run_in_background`), auto-selected
`test-driven-development`	Red-green-refactor discipline for features and bug fixes
`systematic-debugging`	4-phase root cause analysis: gather, hypothesize, verify, fix
`writing-plans`	Create bite-sized task plans with acceptance criteria
`using-git-worktrees`	Create isolated workspaces for parallel development
`brainstorming`	Socratic exploration of requirements and design trade-offs

Hooks

Hook	Event	Function
`git-guard.py`	PreToolUse:Bash	Blocks commits on main/master, blocks push to main/master, warns on raw `gh pr merge`
`task-completed-gate.py`	TaskCompleted	Prevents implementation tasks from being marked complete before review
`stop-check.sh`	Stop	Warns about uncommitted changes, open PRs, changes on main
`workflow-preferences.sh`	SessionStart	Injects execution preferences (use plan-execution)
`inject-current-time.py`	UserPromptSubmit	Injects current local time on every prompt (~12-25 tokens). Optional late-hour and US-DST nudges via env vars — see below

Time-injection hook

inject-current-time.py runs on every user prompt and prepends a bracketed system note like [current time: 2026-05-26 14:04 EDT]. This closes the wall-clock-awareness gap caused by Claude's training cutoff — without it, the model cannot reason reliably about "this morning" vs. "six months ago," whether a release / changelog entry is recent, or judge urgency.

Two optional suffixes, both off by default and opt-in via environment variables (set in your shell rc):

Env var	Effect
`CLAUDE_TIME_LATE_START` + `CLAUDE_TIME_LATE_END`	When both are set to integer hours and the current hour falls in `[start, end)`, the note gets a `past late-hour wrap-up cutoff` suffix. Useful as a self-care nudge for late-night coders. E.g. `CLAUDE_TIME_LATE_START=2 CLAUDE_TIME_LATE_END=5`
`CLAUDE_TIME_DST_REGION=us`	Adds a US DST transition reminder within ±3 days of the spring-forward (2nd Sunday of March) and fall-back (1st Sunday of November) transitions. Only `us` is implemented.

The hook is pure Python stdlib, ~5 ms per invocation. To disable entirely, comment out its entry in hooks/hooks.json after install or override at the user-settings level.

PR Review Verification

Before merging a PR, verify the latest review comment was generated for the current head commit. The plugin's claude-code-review.yml workflow template stamps the head SHA into the comment body (**Reviewed commit:** <short-sha>). Downstream verification:

CI passes — gh pr checks <pr> --watch
SHA match — parse **Reviewed commit:** <short-sha> from the latest claude[bot] sticky comment; compare to gh pr view <pr> --json headRefOid -q .headRefOid. Mismatch / missing / null → wait for next run (retry 3× with 5s backoff for replication lag).
Assess — scan for CRITICAL / FIX / BLOCKER markers.
Verdict — READY TO MERGE or CHANGES NEEDED.

Why SHA match instead of timestamp polling? Timestamp arithmetic is brittle under GitHub's distributed clock skew and silently fails when CI is re-run without a fresh push (the old comment still passes a timestamp filter). The SHA stamp is direct and unambiguous. The review workflow ALWAYS posts a comment — never treat absence or staleness as tacit approval.

The plan-execution skill runs this inline as part of its final fix-loop. Project-specific bits (bot identifier, comment format, repo quirks) belong in each consuming project's CLAUDE.md and per-project memory.

Usage Examples

Checking PR Readiness

> Is my PR ready to merge?
# Claude runs gh pr checks --watch, then SHA-matches the latest sticky review comment
# against PR head, then reports verdict (READY TO MERGE | CHANGES NEEDED)

Merging a PR

> Merge my PR
# Claude verifies readiness (CI + SHA match) first, then merges if READY TO MERGE

Code Review

> Review my recent changes
# Claude invokes staff-code-reviewer agent

Pre-commit Verification

> Verify my code before I commit
# Claude invokes code-verifier agent

Customization

Review Criteria

Add organization-specific review criteria to agents/staff-code-reviewer.md:

### 6. Organization Standards
- [Your coding standards]
- [Your security requirements]

Additional Git Guards

Add patterns to hooks/git-guard.py:

blocked = [
    (r'git\s+push\s+.*\b(main|master)(\s|$)',
     "BLOCKED: Direct push to main/master."),
    # Add your own:
    (r'git\s+push\s+--force',
     "BLOCKED: Force push requires explicit approval."),
]

Requirements

Claude Code CLI
gh CLI (GitHub CLI) for PR operations
Git
For Windows: Git for Windows (provides bash for hooks)

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 24 Commits
.claude-plugin		.claude-plugin
.github/workflows		.github/workflows
.serena		.serena
agents		agents
commands		commands
hooks		hooks
skills		skills
.gitignore		.gitignore
.release-please-manifest.json		.release-please-manifest.json
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
release-please-config.json		release-please-config.json

Folders and files

Latest commit

History

Repository files navigation

Development Workflow Plugin for Claude Code

Features

Philosophy

1. Planning Before Implementation

2. Parallel Execution with Mode Selection

3. Parallelization Through Isolation

Complete Development Workflow

Feature Development Path

Bug Fix Path

Git Guards

Execution Preference

Installation

Option 1: Direct from GitHub (Recommended)

Option 2: From Local Clone

Components

Slash Commands

Agents

Skills

Hooks

Time-injection hook

PR Review Verification

Usage Examples

Checking PR Readiness

Merging a PR

Code Review

Pre-commit Verification

Customization

Review Criteria

Additional Git Guards

Requirements

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 7

Packages 0

Uh oh!

Uh oh!

Contributors

Uh oh!

Languages

Packages