Operations skills

Stand up an quarterly operating system for an agency end to end, the two-page vision and plan document, a seat chart (seats and roles before names), a weekly scorecard of 5-15 measurables, quarterly priorities with owners, a single prioritized issues list, and the weekly leadership meeting agenda that runs them all on cadence.

Classical responsible-AI and algorithmic-fairness review, audit a model or automated decision system for bias and fairness (demographic parity, equalized odds, disparate impact), write model cards, run an algorithmic impact assessment, and classify it under EU AI Act risk tiers with NIST AI RMF and ISO 42001 control-ID mapping.

Clean regenerable app and repo waste (OS junk, logs, coverage, Playwright reports, build caches, empty generated dirs) with a checksummed archive and no silent deletes. Not UI de-slop (as-core-application-deslop), not folder organization (as-ops-folder-cleanup), not disk defrag (as-ops-defrag-compact), not dead-code removal (as-eng-deprecation-migration).

Generate and maintain robust developer and user documentation for the app, populate /docs/dev and /docs/user from the codebase, planning docs, and commit history, and power the in-app docs/support section.

Deep-dive production-readiness audit of the whole app (pages, API routes, UI/UX, security, schema, PRD gaps, placeholders vs real functionality) scored so every category normalises 0-100, a weighted composite, and an honest denominator where an unassessed factor is reported as a gap instead of scored zero.

Run the board meeting machine and keep the governance record clean, deck construction on the narrative arc (performance vs plan → the one decision → asks), pre-reads sent early so meetings decide, not present, annual governance calendar (approvals, comp, audit, budget), minutes and resolutions drafted to counsel-review standard, action tracking, committees, and director onboarding.

Generate, consolidate, and maintain a project's build task queue, turning spec and planning docs into unchecked tasks in its task list (Bootspring format when connected), merging stray queues into one source of truth, and keeping the loop fed.

The step-0 foundation that makes a new venture a legitimate, payment-ready business, a plain-English orientation + checklist for setting up a legal entity (LLC/etc.), an EIN, a business bank account, and payment processing (Stripe/GHL), with a vendor-neutral comparison of formation paths, a business-naming strategy, the order of operations, and the ongoing-compliance budget.

Build a robust, long-format, INDUSTRY-SPECIFIC business plan via a multi-step full-context Q&A, uncover the business one topic at a time (industry first), assemble a tailored long-format outline the owner approves.

Analyze workload vs available capacity and forecast utilization, then recommend automate / deprioritize / hire. Distinct from as-ops-status-report (project status). This is the bandwidth math.

Mine a screenshot library (Snagit, CleanShot, macOS screenshots) for proof by reading each capture's embedded OCR metadata first, so thousands of images become greppable text instead of a manual scroll. Then triage, redact sensitive strings, and publish.

Write a change request with impact analysis, rollback plan, CAB-style approval, and comms. Binds the production-change gate when the change touches the app/infra (owner runs it).

Cut CI/CD and hosting spend (relax or gate GitHub Actions workflows, stop redundant runs, disable costly crons, and tame Vercel auto-deploys) without breaking the safety nets that matter.

Track competitors and produce battlecards and positioning gaps.

Take one competitor name/URL and produce a polished, client-facing competitive teardown, scraped positioning/pricing/offers, a feature + messaging matrix, gap + threat analysis, and a presentation-ready render that shows where the rival's offer stops. Not an internal battlecard: a leave-behind a prospect can read.

Track the compliance program and audit readiness, requirement inventory, control owners, evidence, and gaps for SOC 2, ISO 27001, and GDPR/CCPA. Distinct from as-ops-legal-compliance (reviews a specific contract/policy) and as-sec-* (tests controls). This is the compliance-program tracker.

The comms layer when something goes wrong (outage, data incident, viral complaint, bad press, layoffs) severity, activation, holding statements (what we know / what we don't / next update), stakeholder sequencing (affected before press, staff before public), drafts per channel (status page, email, social, press, internal), cadence contract, one-voice spokesperson law, do-not-say list.

Durable Architecture/Operating Decision Records (ADR/ODR) (the WHY behind a costly-to-reverse choice) context, options considered, the decision, consequences, reversibility (Type-1/Type-2), who decided, and what was rejected, mined from meetings and sessions into an append-only log with a proposed/accepted/superseded lifecycle. The institutional memory an agency loses when contractors churn.

Compact a workspace: flatten needless nested folders, pack bulky generated output into a dated archive, and run git gc only on a clean tree. Not an APFS/Windows disk defragmenter.

Stand up and run a self-hosted document-management archive (Paperless-NGX-class), the tag / correspondent / document-type taxonomy, the OCR → auto-classify → tag → full-text-index pipeline, faceted retrieval, and Restricted-class governance (least-privilege, audit-logged, human-gated retention) on owned hardware.

Run and grow an online store (Shopify/DTC/marketplace), model the unit economics (AOV, COGS, margin, CAC, LTV, contribution margin), write converting product listings + merchandising, manage catalog/inventory + reorder points, set fulfillment + a lawful returns flow, run store-funnel CRO + abandoned-cart recovery, and pair acquisition with an email/SMS retention engine (where DTC profit lives).

The end-of-day report you can actually send, to an employer, a client, or a stakeholder. Turns the day's real work into a short, evidenced, client-safe update covering what was accomplished with proof, what moved, what is blocked and on whom, and what happens tomorrow.

One input (a live site URL + optional repo) → a measured, calibrated 1-100 experience scorecard for the whole web property, marketing site and app. Runs a real tooling harness (Playwright, axe-core, Lighthouse, Core Web Vitals, security-header probes) across desktop/mobile, scores 17 dimensions, calibrates against a best-in-class bar, and renders a branded scorecard + prioritized roadmap.

Run revenue and billing operations, MRR/revenue tracking, failed-payment recovery, refunds, basic reconciliation, and finance reporting.

Question every requirement against physics, rewrite the spec so only the engine, surviving regulation, and owner-confirmed taste remain. Past-defect RCA is as-ops-root-cause; order of work is as-ops-the-algorithm; PRD writing is as-pm-write-spec.

Clean and organize a folder tree (empty directories, duplicate files, broken symlinks, misplaced dumps, and a proposed canonical layout) archive-first, owner-gated source moves. Not app caches (as-ops-app-cleanup), not disk defrag (as-ops-defrag-compact), not planning-doc compaction (as-core-planning-compaction).

Diagnose and run a business as four sequential pillars (Value Creation, Marketing, Sales, Value Delivery) find the ONE binding constraint and pull that pillar. Leaves the standing operating system to as-ops-agency-operating-system, moat/strategy to as-ops-strategic-advisor, and past-defect RCA to as-ops-root-cause.

Package and run fractional CMO/COO/CFO/CTO services as a productized agency line, role-scope matrices (owns vs advises vs never touches), retainer tier math, the 90-day arc (diagnose → quick wins → operating rhythm), per-role artifacts + client scorecard, capacity math, and hire-your-replacement transitions inside counsel-gated liability boundaries.

Prepare and run a capital raise + ongoing investor relations, decide whether/how much to raise and tie use-of-funds to milestones, build honest materials (deck, assumption-backed model, one-pager, a controlled data room), target + track investors through a tight process, get cap-table literacy, and run a candid post-raise investor-update cadence.

Point it at anything (an app, a website, a repository, a document set, a process) and get a complete gap analysis covering what exists, what is missing, what it costs to leave open, and who owns closing it. Domain-agnostic where the specialist audits are domain-locked, and it routes into them when depth is warranted.

Operate a fleet of GoHighLevel sub-accounts via the v2 REST API and productize an agency system into a sellable GHL snapshot.

Extract one shared booster and many vertical fairings from a fleet of duplicate parts. PPL capture extract stays as-ops-lead-gen-tesla-build; this is the general platform move.

Run a 90-day goal sprint that turns an ambitious target into a daily activity number you actually hit, set the quarterly goal, reverse-engineer it to the daily activity from real conversion math (e.g. revenue goal → deals → consults/day at your close rate), make it public with an accountability partner, track the daily number, review weekly, then reset and repeat.

Run the compliance-heavy grant and RFP/RFI/RFQ response machine, bid/no-bid scoring with the courage to no-bid, the compliance matrix before any writing (non-compliance is the #1 disqualifier), fresh response-library reuse, outlines mirrored to evaluator weighting, pink/red/gold reviews, grant eligibility/budget/reporting discipline, receipted submission ops.

Run a general hiring and recruiting system for one role end to end, a role scorecard (mission, outcomes, competencies) before the job post, an on-brand job post and sourcing plan, a screening rubric and funnel, a structured interview scorecard, a paid test task that proves the work, and an offer/decision package.

Operations category hub: routes to the right leaf for running the company (runbook, process, status, decisions, meetings, goals), money and governance (finance, fundraising, board, procurement, vendors, insurance, compliance, risk), people (hiring, people-ops, succession), hard-numbers strategy (first-principles, the-algorithm, make-vs-buy, idiot index, throughput).

Stand up one SSO/IdP in front of every self-hosted admin UI (Authentik default; Authelia or Keycloak as swaps) with OIDC, least-privilege apps, and Tunnel/Access in front. Not Cloudflare Access alone (as-devops-cloudflare), not password vault ops (as-devops-env-secrets-ops).

Compute should-cost vs actual (idiot index) for a part, process, SaaS seat, or deliverable and route delete / redesign / in-house / keep. Vendor TCO is as-ops-vendor-review; the buy process is as-ops-procurement.

Prepare the business-insurance coverage review a licensed broker acts on, plain-language taxonomy (GL, E&O, cyber, workers comp, property, D&O, EPLI, umbrella, media), exposure mapping from the business model, declarations extraction (limits, deductibles, exclusions, endorsements), gap/overlap flags, COI + additional-insured contract compliance, renewal package, claims readiness.

Run investor relations after the money is in (the monthly/quarterly investor-update system (locked metric set, lowlights before wins, bad news first, a cadence that never slips), data-room hygiene with a currency SLA and access log, an investor CRM (who invested what/when, follow-on capacity, intro graph), specific trackable asks, and information-rights/annual-meeting readiness) every figure.

Design a test article, abort criteria, telemetry, and one-variable next flight instead of a waterfall campaign. Full campaign plans stay as-launch-campaign-planner.

Turn raw conversations, chats, notes, and decisions into structured, reusable knowledge filed where the team can find it, a knowledge base in Notion (or the Project Brain / files).

Build and upgrade pay-per-lead / local-service lead-gen sites to SpaceX first-principles reasoning and Tesla Highland build quality, shared capture booster, vertical-specific fairing, panel-gap UI on the **site vehicle**, fail-closed lead API, CAPI, TCPA, prompt-before-delete. Owns the vehicle (the live site + capture path).

Review contracts and policies, run compliance checks, and route documents for e-signature for any sender, any business.

One input (a business name + location) → a complete, branded, client-ready local SEO / Google Business Profile optimization pack, GBP optimization checklist, NAP/citation gap scan, category + local-keyword recommendations, review-response templates, and ready-to-publish local landing copy (via as-seo-vertical-page-builder). The "rank in the map pack" deliverable, end to end.

Workspace maintenance front door. Route cleanup, organize, compact, and defrag jobs to the right leaf. Not UI de-slop (as-core-application-deslop), not planning compaction (as-core-planning-compaction), not tech-debt ranking (as-eng-tech-debt), not dead-code removal (as-eng-deprecation-migration).

Classify parts as engine vs mass and verdict make / buy / delete. Vendor scoring is as-ops-vendor-review.

Make meetings produce outcomes, not just notes, prep a briefing before (agenda, context pulled from the knowledge base + prior notes + relevant research, the decisions the meeting must reach), then after, turn the transcript/notes into a clean recap with decisions, owner-assigned action items + due dates, and open questions, filed to the knowledge base and tracked to done.

Specify a live ops console: leading indicators, abort criteria, owners, daily/weekly go-no-go. Exec dashboards stay as-analytics-exec-dashboard; checklists stay as-ops-runbook.

Operate many clients' OWN/AUTHORIZED ad and social accounts without false cross-linking, authorization + ToS decision tree first, the platform's sanctioned multi-account access preferred over anti-detect; fingerprint-isolated profiles + per-client proxies + vaulted credentials as fallback.

Turn one plain-English automation request into a validated, importable n8n workflow JSON. Pick the platform (n8n by default when self-hostable), start templates-first, assemble the node graph with correct expressions, validate structure + expressions + connections + AI-agent nodes, and emit a deployable workflow.

Grow by node placement, occupancy, and uptime. Not more ads. Generic flywheels stay as-funnel-growth-loops.

Design the post-launch telemetry → patch → rollback loop for a live product or membership. Engineering compound loop is as-eng-compound-loop; pre-scale flights are as-ops-iterative-test-article.

Run the ongoing people/HR function after the hire, performance reviews on role scorecards, market-anchored compensation bands + raise/promotion criteria, a plain-English employee/contractor handbook + policies, time-off/leave tracking, performance-improvement plans, and a clean, kind, secure offboarding.

Design and build high-end, animated marketing pages and dashboard UI for the app (homepage, marketing site, and dashboard) with GSAP, Framer Motion, Aceternity, Radix, shadcn/ui, Tailwind, and motion-primitives, kept consistent entry-to-exit. Boundary: PPL/local-service lead-gen site vehicles (SpaceX algorithm, Tesla panel-gap quality, capture/CAPI) go to as-ops-lead-gen-tesla-build.

One input (a process described in prose) → a complete, hands-off process kit (extracted steps, a Mermaid flowchart with exception branches, a one-Accountable-per-step RACI, an edge-case register, and a numbered SOP a fresh operator runs with zero gaps) assembled into a single client-ready file. Distinct from as-ops-runbook (which WALKS a recurring cadence; this WRITES the procedure once).

Analyze a process and make it faster/leaner, map current state, find the bottleneck, eliminate or automate steps, quantify the win. Distinct from as-ops-process-doc (which documents). This one improves.

Run procurement as an operation, intake to PO, request intake + threshold triage (p-card vs PO vs RFP), sourcing to shortlist, RFP/RFQ construction with the weighted evaluation matrix frozen BEFORE proposals arrive, bid scoring on TCO not sticker price, negotiation prep (BATNA, levers, term targets), approval-ladder + PO design, and the renewal/SaaS-sprawl audit that kills maverick spend.

Orchestrator for taking a large product feature from idea to shipped, end to end (spec, system design, threat model, build, and launch audit) landing the spec in the project.

Diagnose throughput: takt vs actual, WIP, bottleneck, changeover, a 14-day rate plan. Capacity hours stay as-ops-capacity-plan; which pillar to pull stays as-ops-four-pillars.

The State of the Union for a project, a complete, evidenced, scored picture of where everything actually stands right now. Inventories what is built versus planned, what is proven versus asserted, what changed since the last pass, and what is blocked, then ships it as a branded report with an honest denominator.

Run recruiting as a standing multi-role operation, not one-off job posts, headcount intake from the budget/FP&A, channel mix per role family, outbound sourcing sequences, referral mechanics, employer brand, a talent CRM with silver-medalist recycling, funnel instrumentation with leak diagnosis, interviewer calibration, agency management, and candidate-experience SLAs.

Architect the revenue system between marketing, sales, and success, lifecycle-stage models with entry/exit criteria and handoff SLAs, CRM field/pipeline architecture plus the field-sprawl audit, routing design, funnel-math instrumentation, forecast-hygiene stage definitions, stack rationalization, and the revenue-leak audit that finds where handoffs drop and data dies.

Identify, score, and mitigate operational/business risks. Build the risk register (severity × likelihood, mitigation, residual). Distinct from as-sec-threat-model (security/STRIDE) and legal-risk (legal). This is operational/business risk.

Structured root-cause analysis of a KNOWN, already-happened problem, rank defects by Pareto (80/20), drive 5-Whys and a fishbone/Ishikawa diagram to a single evidence-backed root cause, split contributing factors from the true cause, then write corrective + preventive actions (CAPA) closed by a verification-of-fix step. Numbers-forward.

Run the recurring daily/weekly/monthly/quarterly operations checklists and surface what needs the founder.

Numbers-forward simulation for business decisions (scenario architecture (baseline to disruption), decision trees with expected value, Monte Carlo over input distributions with confidence intervals, market-response curves, and constraint / system-dynamics models) each fronted by an honesty gate that states an accuracy tier, runs a named-bias pass (optimism, anchoring, base-rate neglect).

Run an operator-owned headed browser sitting that maps a web app's authenticated private HTTP, on an account the operator owns or may run, into a replayable registry callable via API, MCP, or CLI. Not as-agent-computer-use (pixel GUI).

Turn a discussion transcript or feature brief into a passing SPEC.md, the gate a build must clear before code. One input → goals/non-goals, a numbered decision ledger (D-01..) with rationale and rejected paths, atomic testable requirements (R-01..), a requirement→test verification matrix, and the spec landed in /planning.

Produce a project/ops status report, green/yellow/red headline, KPIs vs target, progress narrative from the tracker, risks, and owned action items. Distinct from as-analytics-weekly-metrics (the business KPI scorecard) and as-design-internal-comms (the format library). This is a project/initiative status.

Brutally honest strategic analysis of the business/product, weaknesses and failure modes, the 20% of actions driving 80% of results, low-competition/high-demand opportunities, what's truly unique and how to 10x the moat, and billion-dollar expansion paths.

Survive the loss of any one key person, bus-factor inventory of who uniquely holds relationships, credentials, knowledge, and authorities; continuity runbooks per critical function (who acts, with what access, within what limits); break-glass design with dual control and audit; knowledge-transfer triage and operator manuals; interim vs permanent succession.

Inventory science and demand planning for ecommerce and manufacturing clients, ABC-XYZ segmentation, safety-stock and reorder-point review, OTIF (on-time-in-full) delivery analysis, and forecast-accuracy review (MAPE, WAPE, bias), where every recommendation states its assumptions and a service-level target instead of a false-precision number.

Design a complete team-member onboarding and training program for one role, a role ramp map, the SOP/training curriculum pulled from existing process docs, a structured 30/60/90-day plan with milestones, competency checks at each gate, an access/tools checklist, and the manager cadence that runs it.

Run Elon's production algorithm on any system, make requirements less dumb, delete (gated), simplify, accelerate, automate last. PPL sites stay as-ops-lead-gen-tesla-build; which business function to pull stays as-ops-four-pillars; past defects stay as-ops-root-cause.

Spec owned uptime checks and an optional public status page for any fleet, Uptime Kuma-class probes, Healthchecks dead-man for crons, ntfy/Gotify for pages. Not incident command (as-devops-incident-runbook), not Sentry triage (as-devops-observability-monitor).

Evaluate a vendor or tool on cost, risk, and TCO, then recommend renew / replace / renegotiate. Distinct from legal:vendor-check (the agreement-status/legal check). This is the business cost/risk evaluation.

Vendor a third-party tool or open-source repo into the app and fully white-label it, strip external licensing/branding, remove calls to third-party servers, rebrand, and make it a unique, dependency-free part of the platform.