Workflow plugins

個人開発 MVP の構想〜OpenSpec 化までのワークフロー (skill はフェーズ番号順)

6 skills

Skills and commands to install and run a pluggable, multi-provider agent development process in any GitHub repo.

3 skills · 3 commands

Little-loop skill family — grills a ticket, gets a phased plan signed off, then implements and verifies one phase (one commit) at a time

6 skills

Collaborative and async design, planning, and review-resolution skills: ideation, confluence-ideation, design-for-review, plan-for-review, resolve-review, and implement-feature.

6 skills

Decide the homes and the rules for a multi-agent project, then enforce them with hooks and CI.

4 skills · 1 MCP server · hooks

Composable workflow skills (code/conv/issue).

4 skills · hooks

Run a GitHub Issue through OpenSpec propose → review → apply → archive → PR with nested Claude agents, gates and rubric-driven human review

2 skills · 3 agents

BMAD story-delivery pipeline: the run-dev-cycle orchestrator skill that drives BMAD's bmad-build loop story by story and answers its human checkpoints, the three bmad-* agents it uses, the Telegram ask/notify scripts for unattended runs, and a setup-telegram skill to configure them safely.

2 skills · 3 agents

Custom skills for Claude Code including Seedpeak, Docx Tweak, Iter Mini, proj-mini, and Skill Forge

5 skills

Run scoped tasks on a local opencode runtime as background subagents. Resumable event streaming, a warm shared runtime, and an MCP surface that also works under Codex CLI.

1 skill · 2 commands · 1 agent · 1 MCP server

Code style rules for Java + Spring Boot + Thymeleaf projects: naming, Java conventions, template and i18n discipline, and test conventions.

4 skills · 1 command

Five skills: mind-reading (expand a terse prompt into the full brief you meant, from a model of your own transcripts), high-agency (own tasks to a verified result), iterative-improvement (criteria-driven adversarial improvement loop), agent-sessions (orchestrate Claude, Codex and Gemini child sessions with verify gates, budgets and best-of-N) and competitive-systems (evidence-driven performance work with a portable dev.py).

5 skills

Stop-time review gate via API Qwen 3.7 Max (OpenAI-compatible). Bloqueia stop quando Qwen identifica problema no último turn.

4 commands · hooks

End-to-end change-shipping toolkit: ship (worktree → implement → commit → push → PR → hand off to pr-watch), pr-watch (monitor CI → fix failures → report ready; never merges). Zero per-repo config — commands and branches detected from the repo.

2 skills · 2 agents · hooks

Planner -> Worker -> Reviewer pipeline where Opus plans, Haiku implements (with local Ollama when available), and Sonnet reviews. Provides the /mmf skill, three sub-agents, and the bundled ollama-local MCP server.

1 skill · 3 agents · 1 MCP server

Cost-aware Fable orchestration with Haiku exploration and Sonnet 5 implementation and verification, enforced by deterministic validation hooks.

1 skill · 3 agents · hooks

De werkwijze van het AI-lead-programma als skills: interview voorbereiden en uitwerken, blueprint maken, en van blueprint naar een eerste workflow in Lleverage (bèta), n8n of Power Automate.

5 skills

Claude Code plugin for ECC Agile platform

1 skill · 4 commands

Guarded multi-agent implementation, review, and repair workflow with strict JSON contracts, changelog tracking, checkpoint commits, and repeated researcher-developer-reviewer loops.

1 skill · 4 agents

Skills: new-project, floor-check, template-update, arsenal, guide. Installs the default plugin set as dependencies. No hooks in the default set.

5 skills

Rational Unified Process lifecycle management — state machine, artifact tracking, and autonomous orchestration for iterative software development

1 skill · 3 commands · 1 agent

Full DDD development workflow for React + TypeScript + Vite: spec clarification, domain design with brand types and companion objects (classes only for Repository), DI/DIP composition root, strict TDD with Vitest, and adversarial review

5 skills

Full DDD development workflow for Angular + TypeScript: spec clarification, domain design with brand types and companion objects (classes only for Repository and Angular components), DI/DIP via InjectionToken, strict TDD, and adversarial review

5 skills

Meetings and correspondence discipline for repositories that keep them as files: a configurable contract linter for meetings/<date>-<slug>/ and series files (the contract is a floor distilled from three repositories — a track may resolve as a directory or a single file, index.md is owed only after the meeting has happened), with a project's own conventions switched on by name (comms.meeting-rules); meeting files split by who reads them — the agenda the owner's, the pitch what the counterpart hears, matched by topic name, with plugin templates for agenda, prep, pitch and index; a drift signal for the generated registry and per-track meeting pointers, written as markdown or Obsidian wikilinks with participant, topic and material columns; a PreToolUse guard that refuses to mark an outgoing letter sent — by creating it so or by an edit — without `sent-proof:` — the sending tool's id or that of what the owner sent by hand, checked with the access layer where it is installed, down to who the send went to, or the owner's word — and a PostToolUse check that names one a shell command left so, and a linted attachment checklist the sender can click; the form of an outgoing draft checked per its declared `channel:` (a subject line for email, a separator before the text to send — configurable per project, drafts only), and a draft declared outside comms/ named; each paragraph of a draft's text on one line, on every channel whose receiver keeps line breaks (all but comms.markdown-channels, intercom by default); a PreToolUse guard that keeps the raw .eml out of comms/ and the meetings tree, on Write and on a copy in Bash; who may write to a person and how to address them — `trust`, `addressed_as` and the From pairs in the HQ's people/, which a project without its own finds by asking the access layer for its HQ; and a pending-item summary — who owes what by when, series queues, unsent drafts, meetings without a transcript. The wording of everything it writes into the project is con

4 skills · hooks

Front door: greet, explain the AUDIT ▸ MODEL ▸ EXTEND flow, list installed plugins, and consolidate readiness. A thin orchestrator that delegates to the specialists.

4 commands · hooks

Three-perspective preflight review skill (MELCHIOR/BALTHASAR/CASPAR personas, parallel Task spawn) for high-stakes changes — front-loads post-hoc optimization insights before walltime / cost / destructive state is committed. Triggers on walltime >= 2h, >= 100M row DML, non-reversible cutover, new pipeline layer, >= $10 spend, or long-poll scripts.

5 skills

Wires the mempalace memory server into Claude Code: three hooks that load palace context, pull the user's writing/coding voice, and nudge the session diary, plus a /match-style skill and a /remember handoff for cross-session continuity. MCP-first, with a mempalace-CLI fallback when the MCP isn't connected, and no auto-mining.

2 skills · 2 commands · hooks

Plan retrospective discipline: a self-contained cycle of marker-drop (exit plan mode), session-start nag, and a retro-authoring skill. Forces a pause after every approved plan to capture what worked, what surprised you, and concrete improvements.

3 skills · 1 command · hooks

Spec-driven change lifecycle: explore, propose, apply, archive

5 skills

Enforcement layer for AI-assisted development: hooks, gates, and evals that make the workflow mandatory, not advisory

4 skills · hooks