baaton-skills

Open-source agent skills. Plug them into Claude Code, Cursor, Gemini CLI, Goose, or any AgentSkills-compatible tool.

Skills aren't prompts. They're structured workflows with quality gates, scoring systems, and feedback loops that make your agent actually good at the job — not just "generate me a LinkedIn post" good. We're talking 9.0/10 scored by 6 weighted expert lenses before it touches your timeline good.

---

marketing/ — The Full Marketing Stack

6 skills. 37 files. One pipeline. Everything from "who's my customer?" to "did that post actually work?"

Strategy → Create → Audit → Repurpose → Analyze → ♻️ Strategy

Most people ask their AI to "write a LinkedIn post." That's like asking a chef to cook without knowing who's eating. These skills fix that.

Skill	What it does	The boring version
strategy/	Figures out who you're selling to before you write a word	ICP, personas, positioning, messaging, competitive analysis
create/	Runs the full show: brief → write → audit → iterate → ship	Pipeline orchestrator with hard gates
write/	Knows that LinkedIn ≠ Twitter ≠ Email ≠ Ads	Platform-specific frameworks + anti-AI rules
audit/	6 experts roast your draft. Minimum 9.0/10 or rewrite.	Paul Graham + storytelling + actionability + hook + builder + reader
repurpose/	1 post → 5-8 pieces. Not copy-paste. Real adaptation.	Atomization + platform adaptation + scheduling
measure/	Tells you what actually worked (spoiler: not what you thought)	Metrics, benchmarks, A/B testing, reporting

The audit is the secret weapon

Your AI-written content sounds AI-written because nobody checks it. Our audit panel catches:

• 30+ poison words (if your post says "delve" or "leverage," it's dead on arrival)
• Structural AI patterns (the "It's not X. It's Y." contrast that every LLM loves)
• 6 weighted expert lenses scoring independently (Paul Graham doesn't care about your hook — that's Nina Ramen's job)
• Binary AI pre-check — 3+ flags and you rewrite. No scoring. No mercy.

The result: content that passes as written by a human who actually builds things.

---

baaton/ — Project Management for Agents

API-first issue tracking via Baaton. Because your agent should be able to create a ticket, not just write code.

---

Install

Claude Code (all marketing skills)

git clone https://github.com/rmzlb/baaton-skills.git /tmp/baaton-skills
for skill in strategy create write audit repurpose measure; do
  cp -r /tmp/baaton-skills/marketing/$skill ~/.claude/skills/$skill
done

Cursor / VS Code

git clone https://github.com/rmzlb/baaton-skills.git /tmp/baaton-skills
for skill in strategy create write audit repurpose measure; do
  cp -r /tmp/baaton-skills/marketing/$skill .claude/skills/$skill
done

Just one skill

cp -r /tmp/baaton-skills/marketing/audit ~/.claude/skills/audit

Each skill works standalone. Install all 6 for the full pipeline, or pick what you need.

---

Quick Start (5 minutes to your first audited post)

You:    /strategy icp
Agent:  [Walks you through ICP definition with templates + examples]

You:    /create linkedin "why your onboarding is probably broken"
Agent:  [Collects brief → drafts → runs 6-expert audit → scores 7.8/10
         → iterates → rescores 9.2/10 → delivers final draft with checklist]

You:    /repurpose linkedin [paste the post]
Agent:  [Atomizes into 7 content atoms → generates X thread + email +
         2 standalone tweets + carousel outline → staggered schedule]

You:    /measure report
Agent:  [Weekly report: best performer, worst performer, pattern spotted,
         recommended next test based on data]

One idea. Six platforms. Scored. Scheduled. Measured.

---

What makes this different from "awesome prompts" repos

Feature	Prompt collection	These skills
Quality gate	❌ Hope for the best	✅ 9.0/10 minimum, 6 expert lenses
Anti-AI detection	❌ "Write like a human"	✅ 30+ poison words, structural pattern matching
Verification loop	❌ One-shot	✅ Max 3 audit loops with evidence checklist
Platform optimization	❌ "Adapt for Twitter"	✅ Algorithm rules, hook rewriting, tone shifts per platform
Strategy upstream	❌ Jump to writing	✅ ICP → Persona → Positioning → Messaging → THEN write
Measurement	❌ Vibes	✅ Metrics, benchmarks, A/B testing, quarterly strategy review
Format	❌ Copy-paste prompts	✅ AgentSkills standard (works with 30+ tools)

---

Repo Structure

baaton-skills/
├── README.md
├── LICENSE
├── baaton/                        # Project management
│   ├── SKILL.md
│   ├── references/
│   └── scripts/
└── marketing/                     # Marketing suite (6 skills)
    ├── strategy/        # ICP, positioning, GTM
    ├── create/            # Pipeline orchestrator
    ├── write/            # Platform frameworks
    ├── audit/             # Expert panel scoring
    ├── repurpose/         # 1→N multiplication
    └── measure/       # Metrics & reporting

---

Contributing

Build a skill that makes agents less mediocre? Open a PR.

Rules: AgentSkills format (SKILL.md + references/), under 500 lines, real examples, tested with at least one agent.

License

MIT. Free forever. Go ship something.