The problem

gstack makes engineering teams ship faster — but the rest of the company is still flying blind. Product launches happen without aligned messaging. Board meetings get ad-hoc tech updates. Nobody knows what the infrastructure actually costs. When it's time for a fundraise, technical due diligence is a scramble.

What this adds

Six skills that extend gstack from "engineering OS" to "company OS" — each with a distinct executive or communications persona:

/cfo — "What does our technology actually cost?"

Maps every infrastructure service to its cost driver. Models scaling projections at 10x and 100x. Quantifies technical debt as a balance sheet (principal + monthly interest in engineering hours). Produces build-vs-buy scorecards. Computes engineering ROI from git history. Answers the question boards always ask: "What's our cost per user, and does it go down at scale?"

/vc — "Would you invest in this technology?"

Technical due diligence from a partner who was an engineer before crossing to the dark side. Moat analysis across 5 dimensions (data, network effects, switching costs, technical complexity, integration depth). Team velocity scorecard from commit history. Architecture scalability assessment. Produces an investment thesis: "I'd write the check because..." or "I'd pass because..."

/board — "Give me the 2-page tech brief for the board meeting"

Single-page KPI dashboard with 4 quadrants (velocity, quality, team, risk). Strategic alignment assessment: are we building the right things, building them right, building fast enough? Technology bets evaluation. Governance checklist. Produces a board-ready brief that a non-technical director can read in 5 minutes.

/media — "What's the story?"

Mines the codebase and git history for newsworthy angles. Crafts launch narratives (headline, lede, proof, vision). Produces press releases, blog post drafts, tweet threads. Incident communication templates (acknowledge, timeline, root cause, remediation). Competitive positioning briefs. Every claim is verified against the actual code.

/comms — "Keep the company aligned"

Internal communications for engineering teams that ship faster than the rest of the org can track. Weekly engineering updates (what shipped, what's blocked, what's next). Incident communications (3-tier: immediate, resolution, post-mortem). RFC summaries for non-technical stakeholders. All-hands prep with slide outlines and speaker notes. Change management templates. New engineer onboarding materials generated from the codebase.

/pr-comms — "Control the narrative"

External PR strategy. Press release drafting (standard format, not corporate fluff). Crisis communication plans with hour-by-hour playbooks. Social media content strategy (Twitter threads, LinkedIn posts, HN titles). Thought leadership content from engineering insights. Media targeting with tiered pitch templates. The rule: every claim must be defensible from the codebase.

How they work

Same gstack pattern — persona, methodology, structured output:

You:   /vc --moat

Claude: MOAT ASSESSMENT
        Data moat:            Emerging
        Network effects:      None
        Switching costs:      High
        Technical complexity: Deep
        Integration depth:    Moderate
        
        OVERALL: Narrow moat
        Defensibility: ~18 months before funded competitor catches up
        
        VERDICT: Architecture is solid but moat is thin. The switching
        costs from deep CRM integration are the real defensibility...

All 6 skills are read-only — analysis only, never modify code.

Why this matters for gstack

Garry built gstack for teams shipping 10K LOC/day. Those teams don't just write code — they fundraise, do board meetings, handle incidents, launch products, and communicate across the org. These skills bring gstack's opinionated, high-rigor approach to the rest of the company's needs.

Integration

• All 6 skills follow the .tmpl → SKILL.md pipeline
• Registered in gen-skill-docs.ts, skill-check.ts, and test files
• LLM-as-Judge evals included with comms complementarity cross-check
• 129 static tests pass, 0 failures

Test plan

• bun test — 129 pass, 0 fail (803 assertions)
• bun run gen:skill-docs --dry-run — all FRESH
• Each skill: valid frontmatter, preamble, structured outputs, read-only
• EVALS=1 bun test test/new-skills-llm-eval.test.ts — LLM quality evals (~$0.14/run)