Each skill is a self-contained, versioned, evaluated workflow that handles a recurring CSM job. Highlighted skills have been contributed upstream to the broader CS organization. Click any skill to see what it does and an example of what it produces.

• csm-morning-brief
• csm-midday
• csm-eod-wrap
• csm-meeting-prep
• csm-meeting-recap
• csm-customer-followup
• csm-account-research
• csm-churn-risk
• csm-email-drafting
• csm-jargon-translate
• qbr-report-generator
• customer-usage-analytics
• csm-changelog-report
• cs-value-report
• csm-account-tracker
• csm-account-enrichment
• csm-account-intelligence-hub
• csm-portfolio-data-sync
• csm-business-reviews-live
• csm-upcoming-meetings
• csm-jira-tracker-hub
• csm-policy-check
• csm-internal-comms
• csm-task-command-center
• csm-tracker
• csm-deploy
• csm-doc-audit
• csm-repo-bootstrap
• csm-scrub
• csm-self-healing
• csm-skill-maintenance
• csm-session-handoff
• csm-session-review
• csm-tooling-watch
• csm-audit-runner
• csm-exec-portfolio-hub
• csm-exec-this-week-hub
• csm-exec-manager-1on1-hub
• csm-meeting-prep-auto
• csm-slack-digest
• csm-slack-ingest
• pre-push-review
• codex

43 of 54 shown. Each skill is git-versioned and follows a maturity progression: draft → tested → trial → crystallized.

Before any usage data, before any product roadmap. The slide that says "I understand what your business is trying to do." Proves the research that earned the seat at the executive table.

Built from the sources every prepared CSM should be using: prior customer conversations, internal account history, recent public announcements, support escalation history, and stakeholder conversations across multiple customer teams. The slide below is a real rendered output from my QBR generator, run against an anonymized profile (customer name and team identifiers replaced with synthetic placeholders; structure, density, and rendering unchanged from production).

Most CSMs open with usage charts. The customer's CFO doesn't care about usage charts. They care whether you understand the business and whether their investment is working. This slide answers both questions in 30 seconds, and is the reason I get invited back to the executive table quarter after quarter.

Stack Custom HTML/CSS slide template · JSON-driven content · Synthesized from public filings + internal research

Stakeholder-aligned outcomes instead of single-thread relationships. $2.5M upsell ARR delivered through pod execution at one role; 100% logo retention across high-risk enterprise accounts at another. The cross-functional motion is the durable lever, not the heroic CSM.

At the enterprise tier, no CSM gets to be a single point of contact. The customer has six departments and the vendor has six functions, and the work is keeping all twelve aligned on the same roadmap. I've stood up two versions of this: a "Rescue Squad" for high-risk accounts in my current role (CSM + AE + SE + RAM operating off a single shared plan), and an account pod model in a previous role that drove $2.5M in upsell ARR by replacing fragmented engagement with a unified cadence.

Beyond the account pod, the work spans the whole vendor org: Finance (reconciled 5+ years of legacy revenue data with the CFO's team for 99.9% ARR accuracy), Product (Customer Advisory Board feedback that shaped roadmap), Engineering (escalations, feature requests), and Marketing (case study development). Senior CS isn't customer-facing. It's everywhere-facing.

Operating cadence Weekly internal pod sync · Monthly customer pod-to-pod · Quarterly executive QBR · Single shared notebook of record

Every conversation builds on the last. Account context survives sessions. Voice consistency persists across every AI interaction. The first time AI tooling actually compounds, instead of starting from zero every morning.

For most CSMs using AI tools, every session starts from a blank slate. Last week's customer conversation, the sponsor's preferences, the team's running concerns, prior commitments. All of it has to be re-established on every prompt. I built a persistent memory architecture that solves this: structured account memory, a voice profile that travels with me across all AI assistants, depth-control protocols, and session handoff patterns that make AI tooling actually compound over time.

What an account memory entry actually looks like (synthetic):

Memory is what separates a useful AI assistant from one you have to babysit. Every customer follow-up email I draft today reads the customer's last interaction, knows the sponsor by name, references prior commitments accurately, and applies my voice consistently. That's not magic. That's deliberate architecture, and it's the layer most CS teams haven't built yet.

Stack JSON-backed account memory · Cloudflare Agent Memory product (dogfooded) · Voice profile in markdown · Session handoff via structured project logs

Cost reduction of 50-65% projected on operational AI spend, with no quality drop on customer-facing output. The architecture work and playbook to make it executable across an entire CS team.

Most AI tooling at the operator level uses a single premium model for everything. That works until the bills arrive. I designed a model-tier routing layer that classifies each skill by its actual reasoning requirement and routes accordingly: cheap models for the structured, deterministic, repetitive work; mid-tier for synthesis and drafting; premium models reserved for customer-facing judgment, voice-sensitive output, and architecture decisions.

AI cost is moving from a small operational expense to a top-five budget line item across Customer Success organizations. The teams that invest in model-tier routing now will have 2-3x more usable AI capacity per dollar than teams that don't. The architecture and policy work to make this real are non-trivial: every skill needs a classified reasoning requirement, every routing rule needs to be verified for quality regression, and the team needs a playbook for when to override defaults. That foundation work is the high-leverage piece, and it's done.

Stack OpenCode model-pinning · Anthropic Claude (Opus, Sonnet, Haiku) · Kimi-K · Gateway-level routing · Per-skill model selection metadata

Designed an executive-ready measurement framework that correlates AI investment with operational outcomes (hours returned, retention impact, expansion correlation, response-time improvements). The answer to "what did AI deliver this quarter?" with numbers a CFO takes seriously.

AI investment without measurement is a cost center waiting to be cut. I designed a structured value-reporting framework that pulls execution telemetry from every skill and automation, correlates it with business outcomes from the CRM, and produces monthly executive reports calibrated for finance and CS leadership audiences. The framework also surfaces underutilized skills (capacity not yet captured) and high-leverage skills (where investment is paying back fastest), turning the report into a planning tool, not just a scorecard.

Across the CS industry, finance teams are starting to ask hard questions about AI spend. The CS leaders who can answer "here's what AI delivered last quarter" with concrete numbers will keep their budgets and grow them. Those who can't will see investment cut. This is measurement work that becomes existential within 12 months, and the framework is the hard part. The reports themselves are the easy output.

Stack Skill execution telemetry · CRM outcome correlation · Monthly executive report generator · Skill maturity scoring

A version-controlled rulebook that AI assistants read at every customer-facing touchpoint. Excerpts from the table of contents:

What used to take 4-6 hours of manual chart work per QBR now compresses by 75%. Across a 44-account enterprise portfolio that's 150-180 hours returned per quarter, or roughly four full working weeks of senior CSM time redirected from spreadsheet wrangling to strategic preparation, executive relationship work, and team coaching.

Most CSMs lose the day before a QBR to data extraction, copy-paste, chart formatting, and slide assembly. The QBR report generator turns that day into a 30-minute pipeline run: telemetry pulled, charts rendered, slides assembled, deck exported as PDF and PPTX. The CSM's job becomes the only thing the customer actually values: framing the story, refining the narrative, anticipating the executive question.

The dollar math is real. At a fully-loaded senior-CSM cost, the time recovered is six-figure annual capacity per CSM, and it compounds across a team. But the bigger impact is qualitative. CSMs who aren't burning the day before a QBR show up to the meeting prepared to think, not just present. That's the difference between a tactical reporting cadence and a strategic partnership.

The charts below are real renders from the generator, anonymized for public display. Data shapes, peak values, and visual treatment are unchanged from production.

Stack Node.js + Puppeteer (headless Chrome) · Custom HTML/CSS chart templates · JSON-driven data model · 14-slide deck output, PDF + PPTX exported

98% logo retention. 44% reactivation of churned ARR ($400K). 70% of the subscription base captured on auto-pay. 99.9% ARR record accuracy in partnership with Finance.

I walked into a Customer Success function with no playbook: no health scoring, no renewal automation, no engagement tracking, no executive reporting. Over 26 months I designed and shipped the Salesforce-native CS operating system that ran 1,200 customer accounts. Custom objects for customer journeys. Formula fields for a 1-10 health score aggregating NPS, CSAT, CES, and engagement data. Automation flows for renewal kickoffs and churn-risk alerts. Executive dashboards in Salesforce + Power BI that gave the C-suite real-time visibility for the first time.

Same operator instinct as the AI builds above, applied with a CRM-native stack and no AI to lean on. Building Customer Success from scratch in a SaaS environment that hasn't run CS before is a genuinely different skill from optimizing an existing CS function. It's the work that earned the 2025 Creative Customer Success Leader Award.

Stack Salesforce (custom objects, formula fields, automation flows) · QuickBooks · DocuSign · Power BI · HubSpot · Pardot

User → Signal collection → Agentic orchestration → Cloudflare infrastructure → Operating surface