A structured, modular knowledge base designed for AI systems, large language models (LLMs), and agent-based workflows. Contains reusable, high-density marketing, sales, and copywriting knowledge optimized for prompt engineering and automation.gned for use with AI systems, agents, and prompt-based workflows.
- Serve as a source of truth for marketing, sales, and growth knowledge
- Provide reusable, LLM-optimized content
- Enable modular usage across tools, chats, and APIs
Core structured knowledge, organized by domain.
- sales/
- marketing/
- copywriting/
- frameworks/
- index.md (complete knowledge index)
Each file:
- Covers one concept
- Is self-contained
- Includes metadata for retrieval and reuse
See @knowledge/index.md for a complete list of all knowledge entries.
Optional extracted modules for direct prompt insertion.
Highly condensed versions of the knowledge base for direct use in chats.
Agent behaviors, instructions, and prompt logic.
Navigate by domain and topic.
- Use individual files as modular context
- Combine multiple files when needed
- Prefer compressed versions when context size is limited
- Atomic knowledge units
- High information density
- Minimal redundancy
- Optimized for LLM consumption
Each knowledge unit must follow a standard structure to ensure consistency and usability in AI systems.
Place files in the appropriate domain folder:
- /knowledge/sales
- /knowledge/marketing
- /knowledge/copywriting
- /knowledge/frameworks
Each file must include:
- Metadata (required)
- Clear title
- Structured content (sections, lists)
Use YAML frontmatter at the top of the file:
title: domain: <sales | marketing | copywriting | frameworks> tags: [tag1, tag2, tag3] use_cases: [use_case1, use_case2] format: <framework | guide | checklist | script>
- One file = one concept
- Keep content concise and dense
- Prefer bullet points over long paragraphs
- Avoid duplication across files
- Optimize for LLM usage (clarity > style)
- Use lowercase
- Use hyphens instead of spaces
Example: b2b-sales-call-script.md
A tiny helper script is available to select an agent spec and an input file with fzf, concatenate them, and copy the result to the macOS clipboard with pbcopy.
Usage:
./scripts/run-agent.shThe script will prompt for a spec and an input file, then copy the combined prompt to your clipboard.
