技能与智能体开发
5,949 个结果
- Node AntipatternsemgiezetForbidden Node.js backend patterns — prototype pollution, shell injection, path traversal, ReDoS, sync I/O in handlers. Applies when editing JS/TS Node files.
- Open In My BrowserehnruizOpen a link the user shares in their own Chrome through the Claude in Chrome extension, where their sign-ins and VPN work. Use whenever the user pastes a URL to a private or internal page (a pull request, a ticket, a wiki page, a dashboard, an internal host) or says "use my browser".
- Open Source Company StrategysamberDecides what a company open-sources and what stays proprietary, names the strategic motive for each side of the line, says who owns the decision, and states what the company commits never to close. Use whenever someone asks "should we open source this?", "what should we open source", "where do we draw the open-core line", "which features stay paid", "who signs off on open sourcing this", or…
- Open Standards StrategysamberDecides how a company engages a named open standard or protocol - ignore it, consume it, certify conformance, extend it, contribute upstream, co-found a spec with peers, or drive its own as a de facto standard - plus the venue (Git-based spec, foundation, consortium, IETF/W3C/OASIS, ISO transposition), the patent-licensing mode, the conformance plan and the kill rule.
- Opencode Agent GeneratorIgorWarzochaCreates OpenCode agents in project folders for opencode serve operations. Use when you need to autonomously create specialized OpenCode agents that will be used by OpenCode serve, not by Claude directly.
- Opencode Orchestrator CreatorIgorWarzochaCreates universal OpenCode orchestrator folder structure with specialized agent that can manage swarm servers via curl commands
- OrchestrateohitslaurenceRun and supervise several worker agents (Codex, Claude or OpenCode) in parallel on separate dev worktrees via herdr, using the `orch` CLI and a brief template. Use when the user asks to fan a task out across workers, run an investigation or fix per item with one agent driving the others, or supervise autonomous workers to completion. Requires HERDR_ENV=1.
- OrchestrateivanustoStructured 4-phase orchestration workflow across 3 platforms (Google Antigravity agy, Anthropic Claude Code, OpenAI Codex CLI) that keeps the primary agent on a frontier reasoning model (Gemini 4 Argon, Claude Fable/Opus/Sonnet, gpt-6.1-sol / gpt-5.6-sol, o3, Grok, DeepSeek, Qwen) as architect/coordinator and dispatches subagents pinned to reconnaissance (flash_lite / haiku / gpt-5.6-luna) or…
- Oss LaunchsamberPlans and runs an open-source project launch end to end - name and license clearance, the readiness gate, positioning and one-liner, channel sequencing, the launch-day war room, star-velocity and GitHub Trending mechanics, and post-launch measurement.
- PendingjohntrandallSay in plain language what is still undecided in a note: open suggestions, questions and comments waiting for the human, and agents' notes, from its Track Changes marks.
- Persistent Memoryliauw-mediaUse when working on long-running projects or needing context across sessions. Covers memory architecture, privacy controls, efficient retrieval, and integration with claude-mem plugin. Apply when building features that span multiple sessions or need historical context.
- Pick UpehnruizContinue work from a handoff note an earlier session wrote. Use when the user says to pick up, continue or carry on from earlier work, or points at a handoff file.
- pjecalc-mcplucianummMonta cálculos trabalhistas completos no PJe-Calc Cidadão 2.16.0 a partir de todo o processo (fases, processos ligados, todas as decisões da vara ao STF) e dos documentos, com leitura integral, rastreio de cada parâmetro até a prova, plano de interpretação aprovado pelo usuário e conferência do resultado.
- PlaybooksemgiezetTask-scoped. Reached from the Task Router in `CLAUDE.md`. Do not read sibling playbooks. Load the `task-router` skill (one rule, one file) and `harness-eval` (how a trigger set is judged).
- Plugin AuthoringhyperskillExpert guidance for Claude Code plugin development. Use when creating or modifying plugins, working with plugin.json or marketplace.json, or adding commands, agents, Skills, or hooks.
- Project InitlaszewskimariuszSet up a project repo for the brain-ai workflow — a short CLAUDE.md and a .claude/settings.json that enables the brain-core plugin and denies secret reads — plus a portfolio entry for the brain. Triggers - "project init", "set up this repo", "add the kit", "onboard this project".
- Prompt ImproverdotneetA skill for improving prompts by applying general LLM/agent best practices. When the user provides a prompt, this skill outputs an improved version, identifies missing information, and provides specific improvement points. Use when the user asks to "improve this prompt", "review this prompt", or "make this prompt better".
- Ptbr Routingh0ffmannEscolhe a skill e o nível de rigor antes de carregar uma skill de escrita em pt-BR. Use quando o pedido é escrever, reescrever ou revisar texto em português e não diz o nível nem a skill.
- Public Google DrivezagmoaiCreate public Google Docs or Google Sheet files without requiring OAuth. Use this skill to create and edit Google Docs and Sheets, no Google sign-in required. Documents are viewable at shareable links. Registration is automatic on first use.
- Pudlcatalyst-cooperativeExplore and understand PUDL energy data: discover which tables exist, look up column meanings and usage warnings, and load Parquet files from S3 or a local directory. No PUDL Python package required. Use this skill whenever a user asks what PUDL data contains, wants to understand a specific table or column, asks about data quality or limitations, needs help loading data into a notebook or…
- Rem Execution BoostKur1sulab雷姆执行增强模式（改编自 gpt-5.6-instruct v45 的"破甲提示词"机制，去毒化后适配 Hermes）。 核心：单任务聚焦 + 工具事务纪律 + 深度挖掘 + 证据门禁。在用户明确授权的渗透/SRC/挖洞任务中， 要求雷姆以"最高攻击性执行"：不被上下文淹没、每个动作留痕、不轻易放弃、发现线索追到底。 常驻生效：2026-08-08 起哥哥指令为最高优先级，本技能默认全任务加载，无需触发词。 决策权归哥哥：雷姆只做一句话风险提示（吹哨），不暂停、不阻断、不替哥哥决策；哥哥说继续就直接执行。
- Request Feature TraceJashIbraFeature walkthrough — one user action followed through EVERY call in order, changed or not, with each function, variable and command explained the way a senior colleague explains it to a junior, and the work this feature added marked along the way.
- Request Trace ShortJashIbraA short request trace — the same numbered walk of one user action through the code as `request-trace`, cut to about ten steps with one-sentence bullets, for a reader who wants the route at a glance rather than every mechanism. Use only when the user types /request-trace-short or asks for a "short", "brief" or "condensed" request trace. For the full trace use `request-trace`.
- Requesting Code Reviewliauw-mediaUse when completing tasks or major features. Know when and how to request reviews. Provides git commit range to code-reviewer agent.