Skip to content

AGENTS.md

Latest commit

 

History

History
275 lines (229 loc) · 13.8 KB

AGENTS.md

File metadata and controls

275 lines (229 loc) · 13.8 KB
╔══════════════════════════════════════════════════════════════════════╗
║                                                                      ║
║    ██╗ ██████╗ ██╗  ██╗    ██╗   ██╗██╗██████╗ ███████╗             ║
║   ███║██╔═████╗╚██╗██╔╝    ██║   ██║██║██╔══██╗██╔════╝            ║
║   ╚██║██║██╔██║ ╚███╔╝     ██║   ██║██║██████╔╝█████╗              ║
║    ██║████╔╝██║ ██╔██╗     ╚██╗ ██╔╝██║██╔══██╗██╔══╝              ║
║    ██║╚██████╔╝██╔╝ ██╗     ╚████╔╝ ██║██████╔╝███████╗            ║
║    ╚═╝ ╚═════╝ ╚═╝  ╚═╝      ╚═══╝  ╚═╝╚═════╝ ╚══════╝           ║
║                                                                      ║
║              M A R K E T E R                                         ║
║                                                                      ║
║    Your Autonomous Marketing Agency — Powered by Agent Skills        ║
║                                                                      ║
║    Developed by 10x.in                                               ║
║                                                                      ║
╚══════════════════════════════════════════════════════════════════════╝

10X Vibe Marketer — Full-Featured Skills Plugin v2.1 — Codex Edition

An autonomous marketing agency powered by specialist agent skills. Give it a marketing task, and it routes to the right specialist to deliver results. Nothing is impossible — the skills handle everything.

First-Time Setup

Run /setup to auto-install all local dependencies (TLDraw SDK, output directories, tracking system).

How It Works

  1. You describe what you need — "Create a content strategy for my SaaS product" or just use a slash command
  2. The 10x CMO analyzes your request — classifies it, identifies the right department
  3. CMO lists assumptions and asks clarifying questions — multiple-choice, never open-ended
  4. Each department agent is available as a skill in .agents/skills/ — any model can execute
  5. Skills execute the task following step-by-step instructions: gather data, create deliverable, save to output/, report back
  6. Quality Judge evaluates everything before delivery (score 0-100)
  7. Deliverables land in output/ — reports, content, strategies, visuals, tracking data

Note: For multi-agent orchestration, use the OpenAI Agents SDK.

Assumption Register (Built-In)

Every agent documents assumptions BEFORE acting:

# Assumption Why It Matters Status
1 [what is assumed] [impact if wrong] CONFIRMED / UNCONFIRMED / FROM PROMPT

This prevents silent failures, wrong decisions, and costly corrections. If an assumption is unconfirmed, the agent notes it as a caveat — never guesses silently.

Model Compatibility

All skills use explicit step-by-step instructions that work across model sizes:

Model Tier Examples Capability
Large (full features) Claude Opus, GPT-4, Sonnet 4.5+ Multi-agent orchestration, complex campaigns
Medium (most features) Claude Sonnet, GPT-4o, Qwen 72B Single-department tasks, 2-3 agent coordination
Small (basic features) Claude Haiku, GPT-4o Mini, Qwen Coder Worker execution, content creation, research

27 Slash Commands

Campaigns & Workflows

Command Description
/campaign Launch structured campaign templates
/strategy Marketing strategy and planning
/audit Run marketing audits (SEO, content, competitor, full)

Department Specialists

Command Description
/seo SEO audits, keywords, on-page optimization
/content Content creation and strategy
/copy Copywriting — headlines, CTAs, landing pages
/social Social media management and posts
/email Email marketing and sequences
/ads Paid advertising — Google/Meta/LinkedIn
/brand Branding and creative direction
/funnel Funnel optimization and CRO
/outreach PR, outreach, link building
/research Market research and analysis
/community Community management

Platform Research

Command Description
/deep-research Multi-platform real-time research (deploys up to 13 platform agents)
/scrape Web scraping — download website assets for analysis
/analyze Content and website analysis with AI
/image-gen Generate marketing images with Gemini AI

Visuals & Tracking

Command Description
/report Generate visual reports
/visualize Create TLDraw canvas visuals
/track View activity tracking and metrics
/status Quick status check

Utilities

Command Description
/setup First-time setup and configuration
/mega Access Mega cloud resources
/help-10x Full documentation and guide
/db Database operations, queries, and management
/schema Review agent schema change proposals

Or just describe what you need in plain English — no commands required.

26 Specialist Agents

13 Marketing Department Agents

Each agent has a unique visual identity in the CLI for easy recognition:

CLI Department Agent Specialty
🔍 SEO seo-specialist Keyword research, technical SEO, audits, on/off-page
✍️ Copywriting copywriter Headlines, body copy, CTAs, landing pages, ad copy
📋 Content Strategy content-strategist Content calendars, pillar/cluster, editorial planning
📱 Social Media social-media-manager Platform strategy, post creation, engagement
📧 Email Marketing email-marketer Sequences, newsletters, segmentation, A/B testing
📢 Outreach & PR outreach-pr-specialist Link building, media pitches, influencer outreach
📊 Research & Analytics research-analyst Market research, competitor analysis, data insights + Research Lead
🎯 Strategy & Planning strategy-planner Campaign strategy, positioning, goal-setting
💰 Paid Ads / PPC paid-ads-specialist Google Ads, Meta Ads, budget planning
🎨 Branding & Creative brand-creative-director Brand voice, visual identity, messaging
🔄 Funnel & CRO funnel-cro-specialist Conversion optimization, landing pages, A/B tests
👥 Community community-manager Community strategy, engagement, moderation
⚖️ Quality Assurance quality-judge Evaluation loop, scoring, quality gate

13 Platform Research Agents

Real-time platform research using WebSearch/WebFetch:

CLI Platform Agent Data Types
🔗 LinkedIn platform-linkedin B2B trends, professional content, company insights
🐦 Twitter/X platform-twitter Real-time trends, public sentiment, hashtags
🟠 Reddit platform-reddit Community discussions, unfiltered opinions
📸 Instagram platform-instagram Visual trends, influencer content, hashtags
👤 Facebook platform-facebook Community groups, ad library, demographic insights
💲 Google Ads platform-google-ads Competitor ad intelligence, ad copy trends
🔎 Google Keywords platform-google-keywords Search trends, keyword opportunities, SERP data
▶️ YouTube platform-youtube Video trends, creator strategies, audience interests
📌 Pinterest platform-pinterest Visual trends, product inspiration, seasonal content
🎵 TikTok platform-tiktok Viral content trends, Gen Z culture, sounds
❓ Quora platform-quora Q&A trends, pain points, knowledge gaps
📝 Medium platform-medium Thought leadership, content formats, publications
🚀 Product Hunt platform-producthunt Product launches, tech trends, startup ecosystem

8 Campaign Templates

  • Full Launch — Complete product/service launch across all channels
  • SEO Audit — Technical + content + competitor SEO analysis
  • Content Calendar — Monthly editorial calendar with content briefs
  • Competitor Analysis — Deep-dive competitor research
  • Email Sequence — Multi-step nurture/sales email sequence
  • Social Media Blitz — Multi-platform social campaign
  • Brand Refresh — Brand identity and messaging overhaul
  • Funnel Optimization — Conversion funnel audit + fixes

Output Structure

output/
├── campaigns/          # Full campaign packages
├── content/            # Blog posts, copy, emails, social posts
│   ├── blog-posts/
│   ├── email-sequences/
│   ├── social-media/
│   ├── ad-copy/
│   └── landing-pages/
├── reports/            # Structured markdown/JSON reports
│   ├── seo-audits/
│   ├── competitor-analysis/
│   └── analytics/
├── research/           # Platform research outputs
│   ├── linkedin/       # LinkedIn research
│   ├── twitter/        # Twitter/X research
│   ├── reddit/         # Reddit research
│   ├── instagram/      # Instagram research
│   ├── facebook/       # Facebook research
│   ├── google-ads/     # Google Ads research
│   ├── google-keywords/# Google Keywords research
│   ├── youtube/        # YouTube research
│   ├── pinterest/      # Pinterest research
│   ├── tiktok/         # TikTok research
│   ├── quora/          # Quora research
│   ├── medium/         # Medium research
│   ├── producthunt/    # Product Hunt research
│   ├── consolidated/   # Cross-platform consolidated reports
│   └── scraped/        # Web scraper downloaded assets
├── images/             # AI-generated marketing images
│   └── generated/      # Gemini AI image outputs
├── strategies/         # Strategy documents and plans
├── visuals/            # TLDraw canvas exports (SVG/PNG)
│   ├── strategy-maps/
│   ├── funnel-diagrams/
│   ├── campaign-flows/
│   └── canvas/
├── tracking/           # Activity tracking and performance data
│   ├── activity-log.md
│   ├── performance/
│   ├── learnings/
│   └── metrics/
├── briefs/             # Content briefs
├── logs/               # System logs
knowledge/              # JSONL knowledge base (persistent intelligence)
├── strategies.jsonl
├── plans.jsonl
├── learnings.jsonl
├── prompts.jsonl
├── errors.jsonl
└── templates.jsonl
db/
├── init/               # Schema DDL and seed functions
├── import/             # Data import scripts (Python)
├── helpers/            # Cross-platform query helpers
└── migrations/         # Schema migrations (001-schema, 002-research-tables)
scripts/research/       # Python research tools
├── web_scraper.py      # Website asset scraper
├── image_generator.py  # Gemini AI image generation
├── knowledge_base.py   # JSONL knowledge base manager
└── config.py           # Research configuration

Tech Stack

See CLAUDE.md for full tech stack details.

Decision Flow

The CMO decides first, then confirms with you:

  1. CMO analyzes your request internally
  2. CMO presents its recommended approach as multiple-choice options
  3. You select your preference (one click)
  4. CMO dispatches the appropriate skill accordingly
  5. All questions use structured multiple-choice (never open-ended text)

Quality Assurance

Every output passes through a Quality Judge evaluation loop:

  1. Department agent creates the deliverable
  2. Quality Judge evaluates with rubric (accuracy, completeness, actionability, relevance, quality, 10x factor)
  3. Score >= 75: Approved and delivered
  4. Score < 75: Sent back for revision (max 2 rounds)
  5. Results logged to tracking system for continuous improvement

Activity Tracking

The system learns from every campaign:

  • What works → output/tracking/learnings/what-works.md
  • What doesn't → output/tracking/learnings/what-doesnt.md
  • Channel performance → output/tracking/performance/channels.md
  • Content performance → output/tracking/performance/content-types.md
  • Strategy effectiveness → output/tracking/performance/strategies.md

Agents read tracking data before starting new work to build on proven strategies.

Codex CLI Notes

  • Skills Directory: All 26 specialist agents are available as skills in .agents/skills/
  • Orchestrator: The 10x CMO orchestrator is at .agents/skills/10x-orchestrator/
  • Configuration: Codex settings in .codex/config.toml
  • Fallback: If AGENTS.md is missing, Codex reads CLAUDE.md via project_doc_fallback_filenames
  • Multi-Agent: Codex CLI runs one skill at a time. For parallel agent orchestration, use the OpenAI Agents SDK
  • Sync: Run npm run sync:codex to regenerate Codex files from .claude/ source

Developed by 10x.in