bengous-plugins

Parallel multi-agent orchestration for complex features with Task-based coordination and git-wt --stack worktree isolation

1 skill · 6 commands · 3 agents

GitHub lifecycle through gh: agent-ready issues, review-ready PRs that show the change in diagrams, images and video, issue and PR triage, CI-gated linear merge, commit-push-PR in one step, stacked PRs with gh stack, and an issue-to-PR loop whose workers land their own low-tier PR

8 skills · hooks

Documentation research through optimal MCP sources and claim verification: research agents, source routing, web and codebase fact-checking

1 skill · 4 commands · 2 agents

Bridge to the OpenAI Codex CLI for cross-model collaboration

6 skills

Understand a topic, a bug, a Claude Code workflow, or the last reply: minimal explanations, root-cause diagnosis, execution traces, layered recaps

5 commands

Code quality and cleanup commands for maintaining clean, maintainable code

2 skills · 2 commands

Local history without an editor: commit with an optional issue-closing trailer, interactive rebase, squash by pattern or hash, GPG re-signing of past commits

4 skills

Sync Claude Code plugin cache from local sources

1 skill · 2 commands · hooks

Opinionated design skills for software excellence: CLI design, system architecture, and more. Skills that push Claude toward intentional, high-quality output instead of generic patterns.

4 skills

Interactive architecture maps in the browser: zones and bricks laid out by ELK, overview arrows routed by libavoid, typed data files checked by a script, and open points discussed in threads between you and the agent; /atlas:board serves a map's page beside your session and hands what you select there to the agent; /atlas:init draws a repository's first map, each claim citing its lines

3 skills

Lifecycle of Claude Code instruction files: audit CLAUDE.md/AGENTS.md/.claude/rules instruction budget, capture session learnings, and resync docs with codebase evolution

3 skills

Plan at the frontiers and review the plan in the browser: decisions and interfaces before mechanics, vertical slices each closed by a check. /vellum:start opens a planning mode of its own: Claude writes the plan and its mockups and grills the reviewer on the open choices; the reviewer comments, edits or approves in a page, and the answer reaches Claude as a prompt.

2 skills · 1 agent

Quality hooks for Claude Code and Codex that a repository opts into with one config file: fixers run on the text before each write, and Stop lanes check only the files the session changed, with a green cache and a loop guard that lets a stuck turn end; in every repository, a landing guard on Bash refuses the one-way doors of a push, a merge or a release; where the profile asks, Haiku refuses once an agent's commit message that tells how the change came about or how it was checked

1 skill · hooks

Meta tooling for Claude Code: write prompts that prompt better (meta-prompt) and aim them higher (ambition)

2 skills

Manage Claude Code settings with JSONC support and schema extraction

2 commands

Audit and clean code comments: protects why/constraint comments, hunts comments that lie

1 skill · 1 agent

Generate 5 unique website redesigns using Theo's pattern: one agent, sequential creation, natural differentiation

1 command · 1 agent

Git branch and worktree cleanup that proves a branch is contained before deleting it and reports what it kept and why: an interactive sweep that asks first, and an unattended one that frees finished agent worktrees and archives abandoned agent branches under refs/archive/

2 skills

Point at your running app and talk to Claude: the session's address serves the app with an overlay that picks an element by a click or a zone by a drag, takes a quick label or a note, and keeps each selection in a batch you send at once; the batch reaches the session as a prompt, each element described by its selector, role, accessible name, box and styles, each zone with an image of it

2 skills

Multi-agent plan review workflow requiring architect and simplifier approval before plan execution

1 command · 1 agent

One-shot repo setup: the dev-trunk/main-release linear model, branches migrated to submodules with GitHub Actions sync

2 skills

Architecture audit that grants the existing code no authority: derive the requirements, design the minimal from-scratch architecture, then classify every component KEEP / SIMPLIFY / REPLACE / DELETE

1 skill

Git worktree helper with stack support for multi-agent orchestration

1 command

Turn a rough task into the completion condition a /goal evaluator can judge: end state, proof command, constraints, blocker, turn cap. Draft-first by default, or interactive clause by clause.

1 skill

A grill where you answer before you see Claude's recommendation: Claude writes each round of questions in its reply, the chat hides every recommendation, and a pane beside it collects your answers and shows a recommendation only when you peek at it

1 skill

Create and edit Mermaid diagrams for software engineering work, architectural ideas, and technical documentation

1 skill

Ship feature branches: strip working files, create PRs, merge to main with GPG signing and squash support

1 skill

Claude asks for a secret in a masked box and gets a placeholder: the value stays in the session's memory, becomes real only inside a Bash command, is hidden again in every Bash and Read result and in the prompt, and never reaches the transcript; the mod writes nothing to disk

Shows the repository's TODOs in a scrolling band above the prompt, newest first: TODO.md items, TODO/FIXME/HACK code comments dated with git blame, and the findings agents record with todos-finding for a batch review; a Mine filter, issue links, picks put in the prompt for Claude, and a toast when one is added