address
01Command Claude Code
Fetch the review and issue comments on a pull request, map each to the exact code it refers to, propose a fix per thread, and draft replies — then apply and respond only after you confirm.
Claude, Cursor, Codex skills and commands
This repository also configures its own agents. See what skills tells them →
Command Claude Code
Fetch the review and issue comments on a pull request, map each to the exact code it refers to, propose a fix per thread, and draft replies — then apply and respond only after you confirm.
Command Claude Code
Drive agent/subagent architecture, configuration, and setup from one command — audit an agent system for failures, check config drift across workspaces, scaffold the .agents/ folder, or wire up dev-loop routing.
Command Claude Code
Name the flagship skill that fits the current situation. Hint only — do not fire other user-invoked skills.
Command Claude Code
Inspect, configure, and reconcile the selected board through one provider-aware entry point. Preserve its existing workflow. The provider may be GitHub, Jira, or another connected system with clearly stated capability limits.
Command Claude Code
Turn a description of something broken into a GitHub issue of type Bug. Drafts a structured report, previews it, and files it only after you confirm — typed Bug where the repo supports issue types, otherwise labelled bug.
Command Claude Code
Clean up what's already done. Default target is git hygiene: verify branches are provably merged into the trunk (squash-merge aware), then prune the merged local/remote branches and stale worktrees they leave behind. Explicit targets extend the sweep to completed GitHub issues and old session files.
Command Claude Code
Claim the next dispatch:codex issue from the GitHub queue and work it end-to-end through codex exec: branch, implement, QA, PR. One invocation handles exactly one task — a local pull loop you trigger, not a daemon. The board/label safety model is identical to /loop; only the engine differs (the local Codex CLI instead…
Command Claude Code
Drive the full deployment and infra lifecycle from one command — ship an app to staging or production, compose a repo-aware deployment workflow, wire an EC2 CI/CD pipeline, configure production monitoring, or scaffold a dev container.
Command Claude Code
Drive the full UI/design lifecycle from one command — run a technical quality audit, evaluate UX holistically, fix layout or copy, tone down an aggressive design, plan a feature before coding, or audit consistency across the app.
Command Claude Code
Strip AI-generated slop while keeping the project buildable. Code slop: console statements, any types, unused imports, dead/commented-out code, redundant comments, unnecessary defensive checks, over-nesting. Product slop (--product): marketing-filler copy, default-shadcn UI, missing loading/error states, dead buttons…
Command Claude Code
Discover the environment variables the code actually reads, keep .env.example truthful, validate the local .env against it, and make sure no secret file can reach git.
Command Claude Code
Capture a rough client or stakeholder requirement, turn it into a PRD epic, split it into scoped sub-issues, and push approved issues to a GitHub Projects kanban.
Command Claude Code
Diagnose the failing GitHub Actions checks on a pull request, find the root cause from the logs, and propose or apply a targeted fix to get back to green — instead of guessing from the red X.
Command Claude Code
Claim the next dispatch:claude issue from the GitHub queue and work it end-to-end: branch, implement, QA, PR. One invocation handles exactly one task — this is a pull loop you trigger, not a daemon.
Command Claude Code
Default = merge them all. /merge reviews every open pull request targeting the trunk, shows you one consolidated plan, and after your yes merges the whole approved set, then reports possible cleanup candidates without pruning them. It is not a single-PR tool and not review-only (that's /merge review).
Command Claude Code
One entry point for the pull-request lifecycle: open or update a PR, review it, digest its comments, tidy it for reviewers, apply review fixes, repair failing CI, or post inline suggestions.
Command Claude Code
Drive the full product-spec lifecycle from one command — create a GitHub issue or local PRD, enforce a spec-first loop, validate completeness, draft a full PRD, intake a stakeholder requirement to a kanban board, or run a discovery interview.
Command Claude Code
Transform a vague prompt into a precision-crafted one — for AI generation, prompt templates in services or pipelines, or system prompts — or debug why an AI response is poor.
Command Claude Code
Run a multi-phase verification pass over completed work — checking for bugs, missed requirements, and incorrect assumptions — before it gets committed. The on-demand version of the QA gate that /loop runs automatically.
Command Claude Code
Point it at what you want improved and it routes to the right engine: strip AI slop, safely refactor, pay down debt, optimize, review structure, or modernize the stack — frontend and backend.
Command Claude Code
Drive the release lifecycle from one command — verify CI and ship, or cut a versioned tag with patch notes — instead of remembering which release skill fits which step. Trunk-based: releases are tags cut from the trunk (the repo's default branch); staging and production are deployment environments driven by CI/CD and…
Command Claude Code
Review whatever you point it at — your working changes, a single PR, every open PR, the last N commits, or a time window — through the same review engine.
Command Claude Code
One entry point for the product-planning chain that turns "who pays us" into a scheduled roadmap: document the ICP, rank the backlog by revenue, then sequence it into GitHub milestones with due dates.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: