The catalog
608,119 rows, by stacks then stars
Lceoliu/codex-fixbook/skills/codex-fixbook Skill
no license0★ repoUse this skill when a user reports Codex CLI or Codex desktop app failures such as hangs, timeouts, missing MCP tools, or MCP initialization problems.
dstevenson1981/jtbd-agent-builder Skill
0★Build agents that replace jobs, starting from a one-sentence business outcome. Use when a user wants to create an agent or skill, automate a job-to-be-done, achieve an outcome ("more sales through digital channels", "cut support cost", "hire faster"), figure out which jobs an agent should do for their business, or set up a self-improving agent with evals. Diagnoses the user's actual system from real data, finds the binding constraint, proposes a dollarized agent portfolio, and emits runnable skills for Claude Code or Codex with metric contracts, graduated autonomy, and a PR-based improvement loop.
tpklo/claude-note/plugins/linked-notes/skills/note Skill
0★ repoCreate linked notes in your Obsidian vault, optionally synced to Notion. Every new note gets wikilinked to its project, to sibling notes in the same folder, and registered in the project's main note — so any future session can discover related context by reading one file. Supports progress logs, plans, client messages, scope docs, research notes, and any project artifact. Use this skill whenever the user types /note, wants to log progress, save session context, capture what happened and what's next, document project status, draft a plan, or create any project-related note. Also triggers on 'note this', 'save progress', 'log this', 'จด', 'บันทึก'.
basejb/likethat/.claude/skills/likethat Skill
0★ repoExtract design DNA from one or more references — LOCAL repo paths and/or web URLs (a site whose mood you like) — understand the CURRENT repo's service, and synthesize a fitting mood — not a clone. Establishes a design system (DESIGN.md + tokens + real code) in the current repo. Use when the user lists local repo paths or reference URLs and wants a service-appropriate look. Triggers on "저 사이트처럼 만들어줘", "make mine like that", "이 레포들 느낌으로 디자인 잡아줘", "이 사이트 무드로 디자인 잡아줘", "참고 레포/URL에서 디자인 시스템", "establish design from these repos/urls". Provide one or more local repo paths or web URLs as arguments.
axtontc/Advanced-Scheduler Skill
0★Schedule background alarms using natural language (e.g. 'in 6 hours', '9:30 AM') that bypass the 15-minute native cap.
jansellmann/decision-interview/skills/decision-interview Skill
0★ repoDecision Interview — acts as your requirements analyst and solution architect, running a structured pre-coding interview that resolves a vague requirement into clear, documented decisions — one question at a time, from problem framing through scope, functional and non-functional clarification, to architecture decisions — producing two output files before any code is written: a `feature.md` for Plan Mode and a `decisions.md` for humans. Use whenever the user wants to plan a new feature, module, integration, or system change, especially when the requirement isn't fully defined. Triggers on: "I want to build X", "we need a feature for Y", "plan this with me", "let's think through Z before coding", "help me architect", or any request to plan rather than implement. Also triggers when the user pastes a vague requirement and asks what to do next. Don't wait to be asked explicitly.
epictaste/claude-eyes/skills/watch Skill
0★ repoWatch a video (URL or local path). Downloads with yt-dlp, extracts auto-scaled frames with ffmpeg, pulls the transcript from captions (or Deepgram transcription fallback), and hands the result to Claude so it can answer questions about what's in the video.
omahdi/astro-guide-skill/skills/astro-guide Skill
0★ repoScaffold and build a distill.pub-inspired Astro documentation guide with data-driven D3.js grid diagrams, built with bun. Single page by default, or a multi-chapter "docbook" with a left sidebar. Use when the user asks to create a technical guide, documentation microsite, design/architecture write-up rendered as a web page, or clean D3/SVG diagrams for docs — especially "distill.pub-style", "guide with diagrams", "docbook / chapter sidebar", or "explain this design as a site".
runsagents/route-ledger Skill
0★Log model-run economics and produce honest, quality-gated routing recommendations. Use when asked to "route models", compare model cost and quality, calculate "model P&L", choose a model tier, audit inference economics, or create a weekly routing report.
Thecsiz/agentic-principles/skills/agentic-critique Skill
no license0★ repoEvaluates, re-storyboards, and designs agentic AI experiences where an agent does work and a human approves, steers, or supervises it. Use for requests such as 'critique this agent flow', 'is this actually agentic', 'is the approval real', 'show me the agent-led version', or 'help me design an agentic flow', and when a user shares a storyboard, screenshot, Figma URL, prototype, or description of an AI feature. Runs two required diagnostics—who is doing the work, and whether approval is meaningful—then scores P0–P15 and checks 13 named anti-patterns. Evaluates agency, oversight, trust, consent, and autonomy; it does not replace usability, accessibility, visual-design, security, legal, or compliance review.
ruide92/proofrail/skills/proofrail Skill
0★ repoEvidence-first guardrail for AI agent maintenance. Use when installing or changing skills, diagnosing repeated agent failures, checking configuration drift, auditing suspicious instructions, or deciding whether a claimed fix is actually verified. Trigger on agent health checks, security reviews, failed automations, skill installation, updates, and before declaring work complete.
dd3ok/onboardkit Skill
0★Use only to initialize, refresh, audit, trim, reconcile, clean, or route active AGENTS.md or coding-agent instructions from local evidence. Never use to review a pull request or code change, even when it changes AGENTS.md; to run the built-in Codex /init command; or for translation, grammar-only edits, release notes, summaries, ADRs, proposals, explanations, or other adjacent deliverables.
linkaka93/jin-yong-name/skills/jin-yong-name Skill
0★ repoUse when the user wants a real Chinese personal name with Jin Yong-inspired naming methods, including newborn naming, adult renaming, sibling or generation naming, surname-based ideation, or Chinese name evaluation. Do NOT use for online handles, fictional characters, wuxia role names, brand names, or purely fortune-based naming.
chuaitry-edu/try-skills/hermes/prompt-architect Skill
no license0★ repoPrompt生成与优化助手。认为大多数烂Prompt不是缺技巧,是没说清楚边界。属于try学习体系的元工具层。
StaryMoon/paper-figure-extractor-skill/skills/paper-figure-extractor Skill
0★ repoExtract figures, captions, page references, contact sheets, and a machine-readable manifest from research PDFs. Use when Codex needs to collect paper visuals for literature notes, presentations, reproduction work, figure comparison, or citation-aware review without manually taking screenshots.
cameronhenkes/skills/skills/user-research Skill
0★ repoEnd-to-end user-research methodology: plan, conduct, transcribe, synthesise and clip a research round. Use when the user says "I want to do user research", "plan a research round", "new study on X", "run a usability test", "synthesise the sessions", "write up the round", "make highlight clips" or "make a sentiment reel". Enforces two gates -- check what prior research and analytics already answer BEFORE designing sessions, and no synthesis without human observer notes. Owns the research plan, discussion guide, mission-route task scripts, the mindset framework, the participant roster, and clip/reel generation with consent-aware captions.
silkyland/transform-my-repo Skill
0★Feasibility analysis and migration strategy for architecture transformations: language ports (PHP to Python, Rust to Go), framework migrations (Vue or React SPA to Next.js), major version upgrades, monolith to microservices and back, database migrations, cloud/container/ serverless re-hosting, sync to event-driven, REST to GraphQL/gRPC, UI platform shifts, standalone to SaaS, and vanilla to framework. Produces an evidence-based GO/PARTIAL/NO-GO verdict with a worth-it test against the do-nothing baseline (dependency census verified in the target ecosystem, semantic gap analysis, per-module difficulty heatmap), then a migration roadmap with behavioral parity gates and rollback points. Allowed to recommend NOT migrating. Use when the user asks to migrate, port, rewrite, re-platform, upgrade a major version, split or merge services, switch databases or API paradigms, asks whether a migration is worth it, or mentions transform-my-repo or /transform-my-repo.
jongall45/frontrun-skills/skills/frontrun-deal-flow Skill
0★ repoSource startups before their round using Frontrun: live pre-round deal signal from the follow graphs of 1,000+ venture investors on X. Use when asked to find early-stage startups, source deals, run a thesis search, check which companies investors are converging on, diligence a startup's momentum, or build a deal-flow pipeline or agent. Works via the Frontrun MCP server or REST API.
nanoodlecom/nanoodle-py/examples/agent-skill/poster-generator Skill
0★ repoGenerate a poster image from a short idea — an LLM expands the idea into a detailed image prompt, then an image model renders it. Use when the user asks for a poster, illustration, or promo image from a one-line concept.
w2478328197-arch/open-health-agent/skills/open-health-agent Skill
0★ repoUse for private personal health and fitness logging, wearable or Google Health/ghealth sync, Excel health ledgers, WeChat text/voice/photo intake, food and nutrient estimates, calorie or TEF calculations, exercise/recovery advice, and persistent health goals. Except when urgent emergency direction must come first, explain the data flow, capabilities, privacy exposure, uncertainty, and non-medical boundary in the user's language on the first use in every conversation.
invisiblebaidu-hue/claude-skill-factory/skills/skill-factory Skill
no license0★ repo多智能体协同开发新的 Claude Code skill。当用户提供 skill 需求文档(路径或内容)并要求开发新 skill 时使用。主智能体只做调度,不读需求/计划/代码/测试报告内容;所有产物通过文件在 1号(skill-planner)、2号(skill-developer)、3号(skill-tester) 子智能体之间传递;同任务连续失败 3 次则停下交人裁决。
xtompie/prism Skill
0★Indirect mode for thinking about a task: it selects lenses (ways of looking) from the base, from the model's own knowledge, and from the internet, and surfaces the challenges, concepts, and decisions to settle BEFORE we build anything. Invoke explicitly.
silverskab/1hbp-ikigai-skill Skill
0★Run a publishing-specific Ikigai discovery session to identify the strongest book topic, reader, positioning angle, and client-attraction path for a coach, consultant, or founder. Use when someone asks what their book should be about, wants to validate a publishing niche, needs a commercially useful book idea, or wants to connect expertise to a book-led offer.
noine32/learn-from-book/skills/learn-from-book Skill
0★ repoプログラミング技術書PDFから技術を1つずつ「実装→テスト→§6客観検証→承認→skill化」で学ぶ体験的学習パイプライン。読むだけで終わらせず、コード実行とミューテーションで"本当に落とし込めたか"を客観検証する。node ランタイム対応(excel-vba は将来)。「本から学んで」「この技術書を検証して」「技術書のこの章を試して」で発動。
mattrmartin/prompt-sharpener/prompt-sharpener Skill
0★ repoTurn a rough, half-formed prompt into a sharpened, high-performing one using Anthropic's own prompt-engineering evidence. Use this whenever the user hands over a draft prompt, pastes a prompt and asks for feedback, describes a task they want to prompt Claude (or another model) to do, or says things like "help me write a prompt", "improve/fix/tighten this prompt", "make this prompt better", "sharpen this", "turn this into a system prompt", or "why isn't this prompt working". Also use when the user is building a system prompt for an app, agent, custom assistant, or GPT. Handles both one-off chat/task prompts and reusable system prompts. Reach for this even when the user doesn't say the word "prompt" but is clearly drafting instructions for a model to follow.
WdBlink/token-firewall-team/skills/token-firewall-team Skill
0★ repoRun Codex Agent Teams through the native subagent lifecycle while preserving bounded contracts, Git truth, deterministic validation, independent verification, and blind final review. Use an external CLI Adapter only when the user explicitly requests a third-party platform such as Claude, MiniMax, or OpenCode.
xiaomihu1992/save-progress-skill/skills/save-progress Skill
0★ repoCreate or refresh a verified CONTINUE.md checkpoint before ending, switching, compacting, or handing off a long Codex or Claude Code session; preserve manual notes and record real repository state, verification, blockers, decisions, and next actions. Do not use for ordinary summaries or status questions. 長い Codex / Claude Code セッションの終了・切り替え・圧縮・引き継ぎ前に、手動メモを保持し、実際のリポジトリ状態、検証結果、課題、判断、次の作業を記録した検証済み CONTINUE.md を作成または更新する。通常の要約や進捗質問には使用しない。
hannsxpeter/auditor-suite/plugins/codeauditor/skills/codeauditor Skill
0★ repoAudit the codebase end to end, write codeaudit.md (a scored, prioritized, self-contained report), then display the results in chat. Read-only: never edits source. Invoke with /codeauditor in Claude Code or $codeauditor in Codex.
ishan5ain/tuiweave-agent-kit/skills/tuiweave-app-builder Skill
0★ repoBuild, migrate, or review Go terminal applications that use tuiweave, with explicit layout, focus, ownership, compatibility, snapshot, scenario, and audit workflows.
DovahkiinYuzuko/somebody-help-me Skill
0★Use this skill when you are stuck on an implementation issue and need to ask another AI for help. It structures the current context, the problem, and your proposed approach into a specific YAML format.
Ludi-Herry/win-encoding-doctor/skills/win-encoding-doctor Skill
0★ repoWindows 中文/非 ASCII 编码问题的分层诊断与修复。症状触发:乱码、中文变 ??、UnicodeEncodeError、0 字节文件、多出 BOM(ef bb bf / ff fe 文件头)、批量替换报成功但文件没变、"在 agent 里正常、换个终端就坏";主动触发:体检/加固 Windows 编码环境、新机器配中文链路。只要用户在 Windows 上报这类症状就用本 skill,别裸猜。负面边界:非 Windows 平台不适用;单纯改字体/显示渲染问题不归这里;写文件必带 -Encoding 之类的日常纪律由 hook 与 AGENTS.md 管,本 skill 管"出事后定位哪层坏了并根治"。
gio-del/skills/skills/content/design-system Skill
0★ repoBootstrap a design system for reveal.js slide decks — interviews you via /grilling for brand identity, colors, typography, and layouts, then generates a theme, named layouts, brand assets, and a brand-guidelines.md. Manual invocation only.
JMSolorzano-13/sdd-skill Skill
0★Specification-Driven Development (SDD) orchestration skill. Use this skill whenever the user mentions .sdd-skill/, state.md, @planner, @architect, @implementer, @review, @tester, @sentinel, SDD, /sdd-skill commands, spec.md in a project context, or wants to continue development across IDEs (Cursor, Claude Code, Antigravity, Open Code). Also trigger when the user wants to start a new project with agentic methodology, adopt SDD in an existing project, manage features/hotfixes with agent roles, or resume work after switching tools. This skill governs a polymorphic agent system where a single AI assumes different roles based on filesystem state — never chat history. Trigger even if the user just says "continue", "resume project", or "what's next" and a .sdd-skill/ folder is likely present.
ReliOptic/campfire Skill
0★원문 또는 구조화 스냅샷을 audience-of-one Campfire 브리프(motion_brief JSON + 재생 가능한 HTML)로 변환한다. 도메인 무관 코어 — 주제는 pack, 숫자는 snapshot이 공급한다.
Databricks data engineering best practices
nijanthan-dev/databricks-data-engineering-best-practices/skills/databricks-data-engineering-best-practices Skill
0★ repoAgent Skill for Databricks data engineering best practices
eric-lim/skills/skills/code-review Skill
0★ repoPerforms a multi-stage code review on git diffs and branch revisions.
sweet-boby/doc-to-skill Skill
no license0★将官方技术文档(Docker、K8s、PostgreSQL 等)提取为结构化知识,自动生成可复用的 Agent Skill。支持 URL/目录/本地文件输入。
ikamal7/fluentcart-dev Skill
no license0★Build, extend, and customize FluentCart — the self-hosted WordPress e-commerce plugin. Use this skill whenever the user is writing PHP that hooks into FluentCart, building a custom payment gateway or add-on, querying FluentCart's database/models, calling its REST API, or asking about FluentCart action/filter hooks, order/subscription lifecycle, products, coupons, shipping, tax, or licensing. Trigger on mentions of FluentCart, `fluent_cart/*` hooks, `FluentCart\App\Models\*` classes, `fct_*` database tables, or `wp-json/fluent-cart/v2` endpoints, even if the user doesn't say "FluentCart docs" explicitly.
ajaygodbole7/java-unit-test-skill Skill
no license0★Write, review, or refactor Java unit tests with JUnit 5, AssertJ, and Mockito. Use when creating test classes for a Java service or component, adding test cases, modernizing or fixing existing tests, building test data (builders/fixtures), or deciding what to mock vs. fake vs. use real. Triggers on JUnit, AssertJ, Mockito, "unit test", *Test.java, @Test, test coverage, or any request to test Java code.
PieroJF/forging-skill/skills/forging Skill
0★ repoUse when preparing a feature, module, or idea before an implementation plan exists — "prepara X antes de construir", "diseña esto", "hazlo airtight" — when it's unclear whether brainstorming or grilling applies, when a spec or decision-log from a previous session exists with unknown quality, or when tempted to patch a defective spec instead of rebuilding it.
yuanyuanyuan430/ganhuo-travel-lock Skill
no license0★Chinese-first travel planning and booking-lock workflow for summer vacation, graduation trips, family trips with children, city recommendations, live ticket/transport/hotel price verification, itinerary lock-in, popular attraction filtering, budgets, pitfalls, and optional Xiaohongshu-style outputs. Use when the user asks for 旅游攻略, 暑假去哪, 毕业旅行, 带娃旅游, 亲子游, 城市推荐, 热门地点避坑, 小红书旅行攻略, 订票, 查票价, 比价, 酒店价格, 高铁/机票班次, 门票预约, 锁定行程, or similar China-focused trip planning.
soltraveler-sri/Steerable-Protocol/skills/eval-authoring Skill
0★ repoAuthor deterministic Steerable eval fixtures for a target integration by deriving cases from its capability registry, validating them against the eval schemas, running them through the target adapter, and emitting coverage plus duplicate-avoidance summaries.
heyallensu/mac-resource-triage Skill
0★Mac first aid for AI coding leftovers: user-invoked vibe-coding triage for memory, CPU, battery, orphan processes, polling configs, and this-session vetted PID kills.
jawsublime-byte/simon-says/skills/battleship Skill
0★ repoRun bounded black-box or blind tests to find failures outside the expected end path without first relying on implementation details. Use only when the user explicitly invokes Battleship for software they own or are authorized to test.
sferarc/pgbeam-skills/pgbeam-connect Skill
0★ repoWire an AI agent to a Postgres database safely with PgBeam. Use this when a task needs the agent to read or query a real Postgres database and you want scoped, read-only-by-default, PII-masked, budgeted, audited, revocable access instead of handing the agent a full-privilege connection string. Covers getting a scoped credential, choosing between a guarded connection string and the hosted MCP endpoint, and paste-ready Claude Code, Cursor, and VS Code config.
xxxxxthhh/disciplined-engineering/skills/disciplined-engineering Skill
0★ repoAdaptable end-to-end software engineering workflow for planning, implementing, refactoring, debugging, reviewing, and finishing non-trivial code changes with proportional process and evidence-based verification. Use when Codex owns a multi-step engineering outcome, especially when a bug needs systematic diagnosis, work may benefit from delegation, or a change must move safely from requirements to delivery.
linxumoney/growth-experiment-planner/skills/growth-experiment-planner Skill
0★ repo判断增长或产品想法是否值得做实验,并生成可证伪假设、指标、护栏、样本与时长评估、停止规则和结果决策。适用于用户提出 A/B 测试、增长实验、转化率优化、版本对比、实验优先级、低流量验证或“这个想法该怎么测”时。
jinzhedev/jz-tanstack-ship Skill
0★Build a complete TanStack Start SaaS application from a product requirements document. 根据 PRD 组装并验证 TanStack Start SaaS;适用于“帮我搭一个 SaaS”、 “用 TanStack 建个站”、“做个带支付的网站”。Guides capability selection (Google OAuth / magic-link login, Stripe subscriptions and one-time purchases, credit ledger), assembles verified reference implementations from assets/, configures services via CLI (Stripe, Drizzle), and generates product pages from DESIGN.md. Use when the user wants to scaffold, bootstrap, or ship a new SaaS site on TanStack Start.
slipalison/claupilot/test/fixtures/basic/.github/skills/pdf-tools Skill
0★ repoExtract text and tables from PDF files using the bundled script
dreamitclean/find-unknowns Skill
0★Finding-Your-Unknowns 方法论(Thariq《A Field Guide to Fable》):用四象限诊断 + 八件仪器(盲点扫描 Blindspot Pass / 多方向原型 / 访谈 Interview / 参照物 References / 可变项置顶的实现计划 / 偏离笔记 implementation-notes / 讲解文档 Pitch / 测验 Quiz) 把任务中的未知转化为明确决策。ONLY use when the user explicitly invokes the methodology or one of its techniques by name — e.g. 盲点扫描 / blindspot pass / 找未知 / 发现未知 / unknown unknowns / 四象限 / 采访我 / interview me / 考考我 / quiz me / find-unknowns / 完整流程走一遍. Do NOT auto-trigger just because a task seems vague, large, or unfamiliar — without an explicit request, handle the task normally without this skill.
yigitkonur/plugin-posthog-cli/skills/posthog Skill
0★ repoThis skill should be used whenever the user asks to inspect, query, debug, audit, or operate PostHog from the terminal: "check our PostHog data", "what are users doing", "investigate this metric", "run a HogQL query", "list feature flags", "find a session replay", "check errors", "manage an endpoint", "upload source maps/dSYMs/ProGuard mappings", "symbolicate this crash", "audit PostHog setup", "is analytics ingestion working", or mentions `posthog-cli`, a PostHog project URL, event names, funnels, trends, retention, persons, recordings, releases, or symbol sets. Prefer this over guessing about product behavior: PostHog contains the real events, identities, sessions, flags, experiments, errors, and release metadata.
Wu030616/quest3-unity-codex-skills/skills/quest3-verify-first Skill
0★ repoVerifies current Meta Quest 3, Horizon OS, Meta XR, OpenXR, Unity package, permission, and device claims against primary sources and the user's actual local project before code or configuration changes. Use for every Quest 3 Unity, MR, Movement SDK, MRUK, passthrough, hand tracking, body tracking, A2E, Android manifest, build, or store task.
hamidettefagh/agent-or-workflow Skill
0★Decides the architecture for an AI feature before it is built, whether the work should be a deterministic workflow, a workflow with a model at the judgment points, a single model call, a single agent with tools, or multi-agent. Use when someone is scoping or designing a new AI use case, asking whether it needs an agent, reviewing a PRD, ticket, or spec for an agent, or pushing back on a design that reaches for an agent where a workflow would do. Also covers Salesforce Agentforce designs, deciding whether a use case should be a Flow, a prompt template, an Agentforce topic with actions, or multiple agents. Reads the design input, extracts the deciding signals with evidence, runs a deterministic verdict, and writes an architecture decision record that biases toward the simplest thing that works.
yuanzhe-jia/data-analytic-skill/skill Skill
0★ repoThinking framework to guide agents through systematic data analysis.
troels-im/agent-skills-playbook-template/.agents/skills/lets Skill
0★ repoUse when a user invokes /lets or asks to start, continue, or inspect a guided development flow such as create-product, create-prototype, continue-work, work-on-bug, improve-codebase, research-decision, or write-docs in an agent-skills-playbook-template repository.
zhoy0409-debug/scholarflow-skills/skills/bioinformatics-workbench Skill
0★ repoUse when a researcher needs to turn a biological question and available sequencing or omics data into a realistic, reproducible bioinformatics workflow with explicit QC gates and deliverables.
uncfreak1255-code/loopspine/skills/loopspine Skill
0★ repoUse for non-trivial software work where the coding agent should plan, test, debug, review, and continue through bounded feedback loops until a real proof gate passes or a named stop condition fires. Especially use when asked to handle work end to end, keep going, work autonomously, use agents or worktrees, fix a regression, prepare a PR, or ship safely. Do not use for simple questions or tiny one-step edits.
hajimi2024/lander Skill
0★帮助 Agent 识别 GitHub 项目的官方使用入口,并完成安装、运行、部署、导入与验证,适用于本地电脑、WSL2 和远程服务器。当用户要求让仓库可运行、可使用、可安装、可部署、可导入,或比较这些官方路径时使用。先判断目标和生命周期,再选择轻量或服务路线,并仅在出现秘密信息、公网入口、既有 Web 栈、脚本管理入口或长期运维时加载对应规则。
labsiqbal/masterplan/skills/masterplan Skill
0★ repoUse when turning a product idea — from a one-sentence thought to a long messy chat — into a complete, execution-ready masterplan package that a coding agent can build end-to-end from a single prompt. Covers idea clarification, prior-art research ending in an absorption map (how heavily to adopt proven prior art), evidence-anchored product interrogation, technical decisions with per-component reference maps, adversarial validation, and revision of existing packages. Also use when the user asks to add features or change scope on a project that already has a masterplan package (revise mode).
deployhq/reply-pr/skills/reply-pr Skill
no license0★ repoAnswer a PR's review comments the right way, in one pass — triage the UNRESOLVED feedback from review bots (Codex, Claude, CodeRabbit, or any other) and human reviewers, then post a reply on each reviewer's own channel in your voice and resolve every thread so nothing is left hanging. Reply-only: it never edits code, and never posts anything without your approval. Use it whenever someone wants to respond to, reply to, answer, or triage the review comments on a PR — even phrased loosely, like "reply to the PR comments", "go through the CodeRabbit and Codex notes and reply", "respond to the reviewers on #1200 and leave nothing unresolved", or "which review comments should I tackle vs skip?". Do NOT use it to edit the code a reviewer flagged — fixing what a reviewer caught is a separate coding task.