A whole marketing department, written as markdown your agent can read

What it does

It’s a library of marketing skills for AI coding agents — each skill is a SKILL.md file with frontmatter triggers and a workflow body, so any agent that reads markdown can use them. Coverage spans marketing plans, campaign briefs, ad copy, competitor research, email, A/B testing, personal branding, AI avatar production, and a dropshipping playbook. Everything is bilingual: a Vietnamese version and a Global mirror (US/EU/SEA/LATAM) of each skill.

The interesting bit

The glue is the idea. There’s no code to speak of — the value is that someone wrote down 63 marketing SOPs in a format agents auto-activate, then chained them into 15 multi-skill workflows (e.g. campaign-launch runs competitor research → insight → plan → brief → content → channels over 14–21 days). A foundation skill (product-marketing-context) is read once per project so the agent stops re-asking about your product every session.

Key highlights

• 63 skills in four clusters: Marketing Core (24 pairs), Personal Branding (7 pairs), a regional flagship (B2B export for VN, dropshipping for Global), and a new Design Master skill covering 8 design types.
• 5 role-based agents (strategist, content producer, performance analyst, channel operator, personal brand builder) that auto-detect VN vs Global mode.
• 15 workflows chaining skills into day-by-day sequences, from client onboarding to a 5-day AI avatar batch.
• Platform-agnostic by design: markdown files, installable into Claude Code, OpenCode, and others; MIT licensed.

Caveats

• The repo description says 138 skills; the README says 63 — the count is fuzzy, so treat the numbers as approximate.
• It’s prompt engineering, not software: quality depends entirely on how well the underlying agent follows the SOPs, and the README’s claims (e.g. “saves ~70% time”, “$2/video”) are unverified.
• Product-building tasks are explicitly out of scope — those live in the companion opa-kit repo.

Verdict

Worth a look if you run marketing for a Vietnamese or global SMB and want your agent to behave like it read the onboarding manual. If you wanted tooling or automation code rather than structured prompts, this isn’t that.