# ai-business-skills, a marketing SOP pack where the trigger is a keyword in the prompt > This pack turns one marketing request into a routed workflow across bilingual Vietnamese and global skill clusters. Here is how the routing works, and where the stated counts disagree with each other. **minhnv0807/ai-business-skills** — 138 bilingual AI marketing skills (69 VN + 69 Global) for Claude Code, OpenCode, Codex, VS Code. Four role SOP packs — content, design, performance, leader ops — plus strategy, personal brand, AI avatar, dropshipping, design master, knowledge library. 4 regions (US/EU/SEA/LATAM) + Vietnam 2025-2026. Companion: opa-kit. - Repository: https://github.com/minhnv0807/ai-business-skills - Website: https://opa.business - Stars: 601 · Forks: 229 - Language: PowerShell - License: MIT - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/minhnv0807-ai-business-skills ## The skill count is stated four different ways Before anything else, the arithmetic does not agree with itself. The repository description says 138 bilingual skills, 69 Vietnamese and 69 global. The README subtitle says 144, the section heading says 144, and the quick start comment says 144 skills copied into the target directory. The system diagram in the same README says 74 in the Vietnamese cluster and 69 in the global one, which is 143. None of these is presented as approximate, so a reader has no way to tell whether two skills were added without updating the description, or whether the description predates a rename. The practical consequence is small but real: you cannot use the advertised number to check that an install completed. ## One `SKILL.md` per job, activated by a keyword in the prompt The unit of the pack is deliberately small. Each skill is a single `SKILL.md` file carrying frontmatter triggers plus a workflow body, and the agent is expected to switch one on by itself when the user's message contains a trigger keyword. That is the entire activation mechanism, and it is the design decision worth judging. It means adding a skill is adding a file rather than editing a router, and it means a skill nobody triggers is dead weight rather than a broken branch. It also means the quality of the pack depends on the trigger vocabulary being distinct enough, which is the failure mode you will hit first: two skills with overlapping triggers and the agent picks the wrong one. ## A foundation skill is read once so the session stops asking One skill sits above the rest and is marked as the prerequisite for everything. The foundation context skill exists so the product, audience, positioning, proof, and objections are read once per project rather than re-requested in every exchange, and the stated saving is roughly 70% of the session's time. The pattern repeats per cluster: each cluster has its own foundation alongside its numbered skills, with a Vietnamese name and a global mirror. That is a sensible design for a pack this size, because the alternative is every one of the hundred-plus skills carrying its own copy of the same product brief and the agent choosing between them. ## Routing runs context first, then cluster, then agent, then workflow The system map describes four hops from a request to an output. A user asks a marketing task, foundation context files are read, a mode router decides between the Vietnamese and the global cluster, six universal agents sit above both clusters, and nineteen multi-skill workflows sit above the agents. The outputs are named as a plan, copy, brief, report, SOP, or dashboard. The reason for the agent and workflow layers above the skills is stated plainly: the skills handle single jobs, and anything spanning multiple steps needs something to sequence them. Nineteen workflows for a single-job layer of this size is the ratio that tells you where the real complexity lives. ## `bash install.sh --global` copies into `~/.claude/skills/marketing/` Installation is a clone and a script: ```bash git clone https://github.com/minhnv0807/ai-business-skills.git cd ai-business-skills bash install.sh --global # 144 skills -> ~/.claude/skills/marketing/ ``` The global flag puts everything under one marketing directory in your Claude skills path rather than scattering it. Windows has a parallel script, and the repository's primary language is recorded as PowerShell, which reflects that rather than the marketing content: ```powershell .\install.ps1 -Global ``` The top level carries two validation scripts as well, one in shell and one in PowerShell, so you can check the pack after installing rather than discovering a malformed skill when a trigger fires. A plugin directory, an MCP configuration, an `llms.txt`, and a separate Vietnamese README are also present. ## Skills are numbered by domain and mirrored one-for-one in English The core cluster is laid out as a three-column table: a Vietnamese slug, a global mirror, and a use case. The first rows read `00-ke-hoach-mkt` against `00-marketing-plan-global` for a fullstack plan, `01-lich-noi-dung` against `01-content-calendar-global`, `05-copy-quang-cao` against `05-ad-copy-global` for six variations across three funnel tiers, and `08-nghien-cuu-doi-thu` against `08-competitor-research-global` for a three-tier competitor analysis with SWOT and a positioning map. The use-case column is where the pack earns its keep, because it tells you what a skill produces rather than what it is about. The numbering is not dense: a second entry-point map routes to skills numbered 30, 31, 32, and a marketing operating system at 34. ## It says plainly that it is not for code, specs, or infrastructure The scope exclusions are as explicit as the inclusions, and three of them are worth quoting rather than paraphrasing. This repository is not for building product code. It is not for writing technical PRDs or specifications, where you should use a separate technical documentation tool. And it is not for deploying production infrastructure, where you should use your own DevOps workflow. Everything else is listed as a use case: full marketing planning for Vietnamese B2B SMEs, global small businesses, and agencies; campaign briefs for TVC, performance ads, and thirty-day content calendars; copy for Facebook and TikTok ads, video scripts, email, and landing page briefs; personal brand strategy and AI avatar production; competitor analysis with a forty-eight hour action plan; a design master covering eight design types through a prompt-driven workflow; dropshipping from niche through fulfilment and scale; and an AI marketing operating system with a brand hub, role-based agents, skill chains, and data loops. ## Releases stop at v3.2.0 while the README news line is at v3.6.0 The release list and the documentation are at different points in the same history. The three most recent tags are v3.0.0 on 2026-05-11, described as a rebrand with a dual edition and three modules, v3.1.0 on 2026-05-24 for a bilingual design master and multi-platform onboarding, and v3.2.0 on 2026-05-25 for Anthropic pattern alignment, marked additive. The news line at the top of the README talks about v3.6.0 completing global parity for four roles and v3.5.0 adding Vietnamese skills, a knowledge folder, a design-producer agent, and four loop workflows. So four minor versions are described without a tag, and the branch itself was last pushed on 2026-09-12. Changelog, roadmap, and version files all sit at the top level, so the history is documented, just not in one place. ## Conclusion This pack suits someone running marketing for a Vietnamese SME or a small global business who wants a repeatable procedure rather than a fresh prompt each time, and who is willing to read the table before installing 144 directories. The foundation-context layer and the one-file-per-job shape are the parts that scale, since they stop a session from re-asking what the product is. It is a poor fit if you need engineering work, since the repository says so explicitly, or if you want a stable count of what you installed. Before you install, run the validation script that ships beside the installer, and reconcile the cluster counts against what lands on disk. ## FAQ ### How do I install ai-business-skills? Clone the repository, then run bash install.sh --global on macOS or Linux, which copies the skills into ~/.claude/skills/marketing/. On Windows use .\install.ps1 -Global. The repository also ships validation scripts in both shell and PowerShell so you can check the pack after installing. ### How does an ai-business-skills skill get activated? Each skill is a single SKILL.md with frontmatter triggers and a workflow body, and the agent switches one on when your message contains a trigger keyword. There is no manual command and no router to configure, so adding a skill means adding a file. ### How many skills does ai-business-skills contain? The published numbers disagree. The repository description says 138, split 69 Vietnamese and 69 global, while the README subtitle, its section heading, and the quick start comment all say 144, and the system diagram shows a 74-skill Vietnamese cluster against 69 global. None of the figures is marked approximate. ### Can ai-business-skills write code or technical specs? No, and the repository says so directly. It excludes building product code, writing technical PRDs or specifications in favour of a separate technical documentation tool, and deploying production infrastructure in favour of your own DevOps workflow. It covers marketing planning, campaign briefs, copy, personal brand, competitor analysis, design, dropshipping, and an AI marketing operating system. ## Sources - [License: MIT](https://github.com/minhnv0807/ai-business-skills/blob/master/LICENSE) - [minhnv0807/ai-business-skills on GitHub](https://github.com/minhnv0807/ai-business-skills) - [Project website](https://opa.business) - [README](https://github.com/minhnv0807/ai-business-skills/blob/master/README.md) - [Releases](https://github.com/minhnv0807/ai-business-skills/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/minhnv0807-ai-business-skills