$ man context-wiki/context-engineering-for-gtm

Context engineering for developers means loading the right files, rules, and constraints so AI writes better code. Context engineering for GTM engineers means loading the right data, voice rules, and workflow instructions so AI produces better campaigns, content, and pipeline operations. The principle is identical. The inputs are different. Instead of code files, you load enrichment data, ICP definitions, voice DNA, anti-slop rules, and campaign history. The agent's output quality is directly proportional to the context quality. Better context, better campaigns.

PATTERN

Layer 1: Data context. What you know about the prospect. Enrichment data, company research, technographics, hiring signals, intent data. This feeds personalization and scoring. Layer 2: Voice context. How you sound. Voice DNA, anti-slop rules, platform playbooks, tone calibration. This feeds content generation and outreach copy. Layer 3: Workflow context. How you operate. CLAUDE.md rules, skill files, pipeline architecture, tool configurations. This feeds agent orchestration and automation. Layer 4: Memory context. What you learned. Lessons from previous campaigns, correction history, handoff files from prior sessions. This feeds continuous improvement. Most GTM teams have layers 1 and 2. The teams that compound have all four.

PRO TIP

Developers use CLAUDE.md to tell the agent how to work on their codebase. GTM engineers use it to tell the agent how to work on their pipeline. What tools to use. What order to run enrichment. How to score leads. How to route to outreach platforms. What voice rules to follow. What mistakes to avoid. Every correction you make to the agent becomes a rule. Every rule makes the next session better. Boris Cherny's Claude Code team does this daily. After every Claude mistake, they add a rule. The system gets smarter not because the model changed, but because the context improved. Same principle applies to your GTM operations.

PATTERN

Prompt engineering: "Write a cold email for a VP of Sales at a SaaS company with 200 employees." Context engineering: Load the enrichment data, the ICP definition, the voice DNA, the campaign performance history, the anti-slop rules, the outreach playbook, and then say "write the email." The prompt is one sentence. The context is everything else. The output quality difference is not incremental. It is categorical. A well-prompted agent with no context produces generic output. A lightly-prompted agent with deep context produces output that sounds like your best operator wrote it. See context engineering vs prompt engineering for the technical breakdown.