The field-kit version of what was taught — the Six-Layer Power Prompt Stack, the move from one-off prompts to compounding skills, and the toolkit you walked out with.
A Power Prompt has six layers. Context, Role, Task, Constraints, Examples, Output Spec. Most people skip the first one — and that's why their AI outputs feel generic.
The compounding move is to save a Power Prompt as a skill. A skill is a file on disk with a trigger phrase. You write it once. It runs forever. The aigtm repo gives you 16 revenue-focused ones plus 45+ more for marketing, ops, and SMB workflows — same install, same repo.
Every skill in the toolkit has all six layers visible in its SKILL.md file. Learn the layers once; read every skill confidently after.
What you know that the AI doesn't. Your ICP, top 3 competitors, differentiator, typical deal size. Written once, travels into every skill.
Who the AI is acting as. Anchors voice, vocabulary, decision framework. Best when the role has a point of view, not just a job title.
One verb, one outcome. The number-one mistake: cramming three tasks into one prompt. Pick one. Chain skills if you need a second.
The guardrails. Where most prompts go wrong by leaving the model to pick the defaults. Length cap, anti-hedging, no-fabrication rule, escalation behavior.
Few-shot anchors. One concrete example beats 500 words of description. The model copies tone, structure, vocabulary, hedge level.
The exact deliverable shape. Headers, tables, length caps, JSON if you're feeding another tool. The difference between "useful" and "drop-in-ready."
A skill is a Power Prompt saved as a file on disk plus a trigger phrase. The six layers travel with the trigger. Reusable. Maintainable. Shareable. You write it once. It runs forever.
The rows below are the revenue-team subset of the aigtm repo — curated for the motions we cover in class today. The repo has 45+ more for marketing, ops, and SMB workflows; browse them at github.com/GTMify/aigtm any time. Trigger phrases are natural language — type them like you'd ask a smart colleague. Customize Layer 1 (Context) once and every skill becomes specific to your company.
| Skill | Trigger phrase | Category |
|---|---|---|
| meeting-prep | "prep me for my call with [company]" | Sales execution |
| prospect-research | "research these prospects" / "find decision-makers at [company]" | Sales execution |
| post-call-summary | "summarize my call" / "draft a follow-up" | Sales execution |
| deal-strategy | "deal strategy for [account]" / "MEDDIC this deal" | Sales execution |
| objection-handler | "this deal is stuck" / "they went dark" | Sales execution |
| pipeline-health | "audit my pipeline" / "pipeline review" | Pipeline + forecasting |
| forecast-narrative | "build my forecast" / "commit email" | Pipeline + forecasting |
| territory-analyzer | "territory analysis" / "rep performance" | Pipeline + forecasting |
| competitive-intel | "competitive intel on [company]" / "monitor competitors" | Market intel |
| win-loss-analyzer | "win/loss analysis" / "why did we lose [deal]" | Market intel |
| qbr-builder | "build a QBR for [client]" | Customer success |
| churn-early-warning | "churn risk" / "at-risk accounts" | Customer success |
| board-update | "write a board update" | Leadership |
| hiring-brief | "hiring brief" / "scope this role" | Leadership |
| weekly-planner | "plan my week" / "weekly priorities" | Productivity |
| inbox-triage | "triage my inbox" / "prioritize these emails" | Productivity |
Pick the path that matches your machine. All three end at the same place — all 61 skills installed and the onboarding flow ready to capture your company context.
Installs Homebrew, Node.js, Python, Claude Code, content tools, and all 61 skills. About 5–7 minutes.
/bin/bash -c "$(curl -fsSL \
https://raw.githubusercontent.com/GTMify/aigtm/main/setup/bootstrap.sh)"Requires winget (built into Windows 11). Clone the repo, run the bootstrap, then reopen your terminal.
git clone https://github.com/GTMify/aigtm.git \
$env:USERPROFILE\claude\aigtm
~\claude\aigtm\setup\bootstrap.ps1Open any skill folder in the repo. Copy the prompt from its COWORK-PROMPT.md file. Paste into Claude Cowork. Fill in the [bracketed fields]. Done.
The class is over when you've turned one of these skills into something that feels like it came from inside your company. Three steps. Under an hour.
~/.claude/CLAUDE.md.
SKILL.md. Edit Layer 1 (Context) — your value proposition, your differentiator, your top competitors. Save. Re-run with the same input. The before/after will sell you on customizing the other 15.
Two questions we surfaced in Session 5. Open in the cohort Slack thread — the discussion is where this gets sharp.
If the six-layer Power Prompt Stack lit something up, the field kit below is the operational layer underneath — how I keep context tokens cheap, configs synced across two machines, and a production safety rail that prevents the worst single outcome.
The moves the live hour didn't have time for: chaining skills into pipelines, custom Layer 1 patterns, anti-hallucination constraints at scale, JSON output for chained pipelines, and when NOT to use a skill. For the prompters who already feel fluent.
A working playbook for treating context tokens like cap-table. Decision tree, reference implementations, shell shortcuts, two-machine sync, memory architecture, production safety rail. Pavilion-branded, ~10-minute read.
A brand-customizable single-file HTML design system from the toolkit. Fill in seven brand decisions; download a self-contained file you can share with your team. Three customization paths: a web form, a Cowork prompt, or a Claude Code one-liner.
Sixteen revenue-team skills, built specifically for the motions you already run — meeting prep, deal strategy, pipeline health, QBR builder, board update. Plus 45+ more for marketing, ops, and SMB workflows. One install. Customizable in minutes.
github.com/GTMify/aigtm →