Lukehle
UserCorporate finance automation skills for Claude Code — month-end close, reconciliation, flux analysis, three-statement integrity, SaaS metrics, NetSuite/BigQuery pipelines, financial artifacts, and measured token economics. Built to run on a managed Enterprise seat.
Categories
Indexed Skills (56)
close-orchestrator
Run month-end, quarter-end, or year-end close as a resumable, ledger-driven process - a durable progress ledger that survives context loss, dependency-aware task ordering, evidence captured per task, and postings staged for human approval rather than executed. Use when starting, resuming, monitoring, or designing a close. Trigger on "close", "month-end", "start the close", "where are we in close", "close checklist", "close calendar", "resume the close", "day 3 of close", or any request to coordinate multi-step period-end work.
context-durability
Survive context compaction during long finance work - checkpoint state to disk before it can be lost, detect that a compaction has already happened, and recover without silently re-deriving a different number. Use before any long-running reconciliation, multi-account flux, large model audit, or multi-day close, and immediately whenever you suspect the conversation has been summarized. Trigger on "autocompaction", "context limit", "did I already check", "where was I in the rec", "resume the close", or any task that will iterate over many accounts, files, or periods.
data-quality-gate
Validate a dataset at the boundary before anything downstream consumes it - control totals, row counts, duplicate keys, referential integrity, period completeness, null and sign sanity, and drift against the prior run. Blocks the pipeline on failure rather than passing bad data forward. Use on every extract, load, or handoff between systems. Trigger on "data quality", "validate the extract", "check the data", "the numbers look wrong", "bad data", "did the load work", "DQ check", "sanity check the file".
fin-artifact
Build financial dashboards and reporting artifacts backed by BigQuery, Google Sheets, or a warehouse extract - in two tiers, an as-of snapshot by default and a live connector mode when the organization exposes one. Carries the finance visualization conventions - ARR and cash bridges, cohort heatmaps, actual-vs-budget variance bands, sign conventions, benchmark bands, suppression rules. Use when building any chart, dashboard, board visual, or data-backed page for financial data. Trigger on "dashboard", "build a chart", "visualize", "live board pack", "board visual", "EBITDA waterfall", "bridge chart", "cohort chart", "financial dashboard", "wire it to the warehouse".
finance-guardrails
The non-negotiable operating rails for any finance automation - the human-approval boundary on irreversible actions, single-writer-per-zone, as-of pinning, materiality thresholds, PII and credential deny lists, and the immutable run log. Load this before automating anything that touches a ledger, a reconciliation, a filing, a board deliverable, or customer/employee financial data. Every other closeloop skill defers to this one. Trigger on "automate", "post", "submit", "publish the numbers", "send the pack", "can Claude just do X", or any request that would change a system of record.
finance-quorum
Adversarial multi-perspective review of a number, model, automation, or deliverable before a human relies on it - independent reviewers each trying to break it from a different angle, then a ruling. Degrades across three tiers so it works on a fully-featured harness or a locked-down Enterprise seat. Use as the review gate before anything is reported, staged for posting, sent, or published. Trigger on "review this", "check my work", "before I send this", "second opinion on the numbers", "defensible to audit", "stress test the model", "what could be wrong here".
flux-analysis
Variance and flux analysis with driver decomposition - month-over-month, quarter-over-quarter, year-over-year, and actual-vs-budget/forecast - producing commentary that cites the specific transactions and drivers behind each swing rather than restating the arithmetic. Use for close review, board prep, budget-vs-actual reviews, and any "why did this move" question. Trigger on "flux", "variance", "why did X change", "budget vs actual", "BvA", "explain the movement", "MoM", "QoQ", "bridge the variance", "walk me from X to Y".
model-audit
Audit a spreadsheet or financial model for correctness defects - hardcoded constants inside formula rows, inconsistent formulas across a row, broken references, circular dependencies, sign-convention violations, unprotected inputs, and error values. Runs deterministic mechanical checks first via openpyxl, then reasoned review of what the machine cannot see. Use before trusting, inheriting, shipping, or building on any workbook. Trigger on "audit this model", "check this spreadsheet", "review the model", "is this model right", "inherited a workbook", "formula errors", "model integrity", "xlsx review".
netsuite
Work with NetSuite as a finance data source and system of record - SuiteQL and saved-search extraction, period/subsidiary/book structure, the transaction and account model, journal entry staging (never posting), revenue and deferred schedules, and warehouse replication patterns. Use when pulling data from NetSuite, reconciling to it, preparing entries for it, or debugging why a NetSuite figure disagrees with a report. Trigger on "NetSuite", "SuiteQL", "saved search", "NetSuite extract", "post a JE", "subsidiary", "accounting period", "why doesn't NetSuite match".
notion-publish
Publish finance runbooks, close documentation, metric definitions, and change records into Notion under a stable single-writer zone contract, so automated updates never destroy human edits. Use when documenting a process, publishing a runbook, mirroring a close summary, or maintaining a definitions page that both people and automations touch. Trigger on "Notion", "publish the runbook", "document this process", "update the wiki", "close documentation", "metric definitions page".
pipeline-change-control
Change management for finance automations that an auditor can read - what changed, who approved it, what evidence exists, how to roll it back, and the run log that proves it. Covers threshold changes, gate overrides, new automations, decommissioning, and stakeholder rollout. Use before changing any automation that produces a reported number, and whenever a data-quality BLOCK is overridden. Trigger on "change the pipeline", "update the automation", "override the gate", "adjust the threshold", "deploy the change", "SOX", "audit trail", "who approved", "roll out to the team".
reconciliation
Build and run account reconciliations - bank, AR/AP subledger to GL, deferred revenue, prepaid and accrual schedules, intercompany, and clearing accounts. Covers matching rules, tolerance bands, exception triage, roll-forward proof, and the aging of unresolved items. Use when reconciling any account, designing a recurring rec, or investigating why a rec does not clear. Trigger on "reconcile", "reconciliation", "rec", "doesn't tie", "subledger to GL", "bank rec", "clearing account", "unreconciled items", "roll forward".
saas-metrics
Canonical SaaS metric definitions, stage-aware benchmarks, and the ARR-to-GL-revenue reconciliation that dashboards silently get wrong. Covers ARR/MRR and the ARR bridge, NRR and GRR, logo vs revenue churn, CAC payback, LTV:CAC, magic number, Rule of 40, burn multiple, runway, and cohort retention. Use when defining, computing, reviewing, or reporting any SaaS operating metric. Trigger on "ARR", "MRR", "NRR", "GRR", "churn", "CAC", "LTV", "magic number", "Rule of 40", "burn multiple", "runway", "cohort", "SaaS metrics", "why doesn't ARR match revenue".
sheets-bridge
Treat Google Sheets (or Excel) as a real finance interface rather than a dumping ground - schema pinning, single-writer-per-range zone contracts, never overwriting human-entered cells, versioned snapshots, and safe pull/push patterns. Use when reading from or writing to a spreadsheet that humans also edit. Trigger on "Google Sheets", "push to the sheet", "update the tab", "read the spreadsheet", "the sheet broke", "someone overwrote", "sync to sheets", "export to Excel".
startup-board-pack
Assemble board, investor-update, and exec reporting packs for an early-stage or growth-stage SaaS company - the narrative arc a board actually reads, cash and runway framing, the metric set by stage, cohort and bridge storytelling, and how to present a miss. Use when building a board deck, monthly investor update, exec review pack, or QBR. Trigger on "board pack", "board deck", "investor update", "board meeting", "exec review", "QBR", "monthly reporting package", "what do I show the board".
three-statement
Balance sheet, income statement, and cash flow statement integrity - the balance check, indirect cash flow tie to balance-sheet deltas, retained-earnings roll-forward, working-capital linkages, and the standard breakpoints where a three-statement model or a reported set of financials stops articulating. Use when building, reviewing, or debugging financial statements or a three-statement model, and whenever the statements must foot before anything downstream uses them. Trigger on "balance sheet", "income statement", "P&L", "cash flow statement", "doesn't balance", "three-statement", "statements don't tie", "retained earnings", "working capital".
tie-out
Prove a number end-to-end before anyone relies on it - source, transformation, control total, agreement to the general ledger or an external statement, with variance measured against a stated threshold. Use whenever a figure will be shown to a human, published, sent, or used for a decision, and whenever a query, extract, script, or model "ran successfully". Running is not proving. Trigger on "tie out", "does this number tie", "reconcile to GL", "check the numbers", "is this right", "verify the report", or before any deliverable leaves your hands.
token-economics
Cost and context discipline for running Claude Code on a managed Enterprise or Team seat - model routing under policy restrictions, cache-TTL discipline, context hygiene, subagent isolation, MCP tool-bloat trimming, and the data-tier hierarchy that keeps large financial extracts out of context entirely. Use when work is slow, expensive, hitting limits, or about to process a large dataset. Trigger on "token", "context", "expensive", "hitting limits", "usage limit", "slow session", "too many tokens", "cost", "optimize the session", "large extract".
usage-audit
Measure actual Claude Code usage from local session data and report concrete numbers - largest token sinks, which tools and files dominate, session and message counts, MCP overhead, and the specific changes that would cut the most. Produces measurements, never generic efficiency advice. Use before optimizing anything, when hitting usage limits, or to find out where the cost actually goes. Trigger on "token audit", "where are my tokens going", "usage audit", "why am I hitting limits", "measure my usage", "what's expensive", "/token-audit".
code-quorum
Adversarial review of a change, a pull request, or a subsystem - lenses for correctness under adversarial input, failure modes, interface and migration risk, operability, and test honesty, with an evidence contract that demands a reproducing case rather than a code smell. Trigger on "review this PR", "review the code", "before I merge", "is this safe to ship", "code review", "audit this subsystem", "what could break".
decision-quorum
Stress-test a decision before committing - surfacing the unstated assumption, the option not considered, the reversibility question, and the failure the decision creates rather than solves. Judges are the constraints that would actually kill it. Trigger on "should we", "help me decide", "which option", "before we commit", "stress test this decision", "is this the right call", "pros and cons", "make the case against".
design-quorum
Adversarial review of a UI, chart, dashboard, or published page before anyone relies on it - lenses for misreading, degradation, comprehension, craft, and data integrity, because design failures are exactly the class the person who built the thing cannot see. Trigger on "review this design", "review the chart", "before this ships", "does this mislead", "design review", "UI review", "check this dashboard", "is this chart honest".
quorum-degradation
Run a quorum on a restricted harness - three tiers from parallel subagents down to a structured single-context pass, what each tier costs in rigour, and how to state the tier so a reader knows how strong the gate that passed their work actually was. Trigger on "no Workflow tool", "can't spawn agents", "restricted seat", "single context", "quorum without subagents", "managed plan", "how do I run this here".
quorum-judging
The judge panel and the synthesis that follows - naming judges as hard constraints rather than extra lenses, when zero judges is correct, consensus and dissent handling, and the ruling format. Covers the one step that survived every run in the corpus - an explicit dissent-honoring convergence. Trigger on "judge", "panel", "who decides", "synthesize", "the ruling", "disagreement", "conflicting recommendations", "prioritize the findings".
quorum-lenses
Coin the lenses for a quorum - deriving them from the subject's own failure and value axes rather than reusing a generic roster, sizing the panel, testing lenses for orthogonality, and the starting vocabulary of seats that have earned their place. Use when setting up any quorum, and when a run returned thin or redundant findings. Trigger on "what lenses", "who should review this", "roster", "perspectives", "the quorum found nothing", "reviewers keep finding the same thing", "set up the panel".
quorum-verification
The adversarial verify phase - the measured differentiator of a quorum. Covers writing refutation prompts that genuinely attack a finding, what counts as reproducing evidence, verdict tallying rules, why abstentions are not support and ties fail, and reading survival rates. Use whenever findings must be proven before anyone acts on them. Trigger on "verify the findings", "refute", "is this finding real", "adversarial check", "prove it", "the reviewers agreed too easily", "survival rate".
quorum
Run a multi-agent quorum - parallel evidence miners with per-subject lenses, adversarial verification, a judge panel, and dissent-honoring synthesis - when an answer must be defensible rather than plausible. Handles audits, gap analyses, post-mortems, architecture rulings, and gated re-runs. Trigger on "run the quorum", "/quorum", "requorum", "continue the loop until the quorum passes", "adversarial review", "multi-perspective analysis", "stress test this", "is this defensible", "second opinion", or any request for a verified analysis. Pass the question or scope as args.
research-quorum
Verified research - delegating the mining to a research capability, then running verification and judging over its output rather than duplicating the search. Covers source verification, the citation-resolves check, distinguishing consensus from repetition, and reporting confidence honestly. Trigger on "research", "what's the landscape", "find out about", "is this claim true", "verify these sources", "literature review", "market scan", "check this research".
mece-decompose
Decompose a question or validate an existing breakdown into branches that are mutually exclusive and collectively exhaustive - issue trees with evidence-coded leaves, kill-test prioritization, and a testability check per branch. Use before dispatching quorum miners, when reviewing a spec or plan for completeness, or whenever coverage matters more than speed. Trigger on "break this down", "is this complete", "MECE", "issue tree", "what are we missing", "decompose the question", "structure the problem". Do NOT use for creative brainstorming, narrative writing, or early exploratory research - premature structure kills those.
app-interaction-patterns
Interaction design for data apps - filtering, cross-filtering, drill-down, brushing, linked views, selection, and the complete state machine every data surface needs (loading, empty, partial, stale, error, no-access, filtered-to-nothing). Trigger on "filter", "drill down", "interactive dashboard", "click a bar", "loading state", "empty state", "error state", "linked charts", "cross-filter", "make it interactive".
artifact-accessibility
Make charts and data apps usable by everyone - colourblind-safe and greyscale-safe encoding, contrast, keyboard operation, focus management, screen-reader structure for SVG charts, data-table fallbacks, and reduced motion. Trigger on "accessibility", "a11y", "colorblind", "screen reader", "keyboard navigation", "contrast", "ARIA", "can everyone read this", "greyscale", "prefers-reduced-motion".
artifact-architecture
The foundational constraints and structure of a published Claude artifact - the strict CSP that blocks every external host, the self-contained requirement, the snapshot-versus-live data decision, page budget, title and favicon discipline, and redeploy semantics. Read this before building any artifact, app, or dashboard. Trigger on "artifact", "build a dashboard", "publish a page", "make an app", "interactive page", "why won't my chart library load", "CSP", "self-contained".
artifact-data-loading
Fit data into a self-contained page - pre-aggregating at the source, choosing a grain, embedding formats and their real byte costs, precision trimming, pagination versus virtualization, and offering a download for the full detail. Use when the data will not fit, the page is slow to open, or you are about to embed raw rows. Trigger on "too much data", "page is huge", "embed the data", "how do I get the rows in", "pagination", "virtualize", "download CSV", "aggregate".
artifact-performance
Make a data page open fast and stay responsive - render budgets, the SVG-to-canvas threshold, event delegation instead of per-element listeners, virtualization, debouncing and rAF batching, avoiding layout thrash, and measuring rather than guessing. Trigger on "slow page", "janky", "laggy", "takes ages to open", "too many elements", "performance", "canvas vs svg", "optimize the page", "scroll stutter".
artifact-testing
Verify an artifact before publishing - the self-contained check, all three theme states, greyscale and print, responsive widths, every data state, keyboard and screen-reader passes, and the observe-before-you-publish rule for connector calls. Publishing distributes, so this is the gate before it. Trigger on "test the artifact", "before I publish", "check the page", "did I miss anything", "pre-publish", "verify the artifact", "review my artifact", "QA".
artifact-theming
Make a page correct in all three theme states - explicit dark, explicit light, and the system default that stamps nothing - plus print and greyscale output for decks that get printed. Covers the guarded-override pattern, the transparent-body bug, chart colours under theme change, and verification. Trigger on "dark mode", "light mode", "theme", "prefers-color-scheme", "page looks wrong in dark", "print stylesheet", "greyscale", "colors invert".
bridge-charts
Waterfall and bridge charts in depth - ARR, revenue, EBITDA, cash, and headcount bridges; anchors, deltas and subtotals; multi-level and stacked bridges; connector lines, sign encoding, label placement, and the footing rule that a bridge which does not sum is a wrong analysis rather than a chart to adjust. Trigger on "waterfall", "bridge", "walk", "how did we get from", "ARR bridge", "cash bridge", "EBITDA bridge", "explain the change", "variance walk", "roll forward chart".
chart-annotation
Titles, labels, reference lines, callouts, legends, and provenance on a chart - writing the takeaway rather than the subject, direct labelling instead of legends, benchmark bands with their stage and vintage, and the footnote discipline that makes a figure checkable. Trigger on "chart title", "caption", "legend", "annotate", "label the chart", "reference line", "benchmark line", "footnote", "what should the title say", "callout".
chart-selection
Choose the right chart for a financial question - a decision table from question to form, the comparison-type heuristic, when a table or a single number beats any chart, and the substitutes for forms that mislead. Start here before building any chart. Trigger on "what chart should I use", "how do I show this", "best way to visualize", "pie or bar", "which chart type", "chart for", "visualize this data".
cohort-charts
Retention and cohort visualization - triangle heatmaps, retention curves, layer-cake revenue charts, and survival views; cohort construction, the fixed-denominator rule, small-cohort suppression, immature-period marking, and reading the diagonal versus the row. Trigger on "cohort view", "retention chart", "churn curve", "heatmap", "layer cake", "NRR by cohort", "survival", "retention by month", "triangle chart".
data-typography
Typography for numbers and data-dense interfaces - tabular figures, alignment, precision and rounding, units, negative-number conventions, type scale, and the label/value pairing that makes a table or dashboard readable at a glance. Trigger on "typography", "font", "numbers don't line up", "type scale", "how many decimals", "format the numbers", "table formatting", "tabular figures", "text is hard to read".
design-tokens
Build a small token system for a page or a pack - colour roles, spacing and type scales, radii, elevation, and chart-specific tokens (categorical series, sequential ramps, semantic up/down) so every artifact reads as one system instead of a series of one-offs. Trigger on "design system", "tokens", "CSS variables", "palette", "colour scheme", "consistent styling", "series colours", "make these pages match".
financial-tables
Table design for financial data - hierarchical P&L and balance sheet layouts with roll-ups, column ordering, conditional formatting that earns its ink, in-table sparklines and bars, totals discipline, sticky headers, and print behaviour. The table is often the real deliverable, not the chart. Trigger on "table", "P&L layout", "financial statement table", "data table", "sparkline in table", "conditional formatting", "totals row", "sticky header", "make this table readable".
flow-charts
Composition and flow visualization - Sankey diagrams for cash flow and customer movement, marimekko for two-dimensional revenue mix, stacked versus 100% stacked rules, and where treemaps and pies genuinely apply. Trigger on "Sankey", "flow diagram", "where does the money go", "customer movement", "marimekko", "mekko", "revenue mix", "stacked bar", "100% stacked", "treemap", "composition".
forms-and-controls
Input design for data apps - period and date-range pickers, numeric and currency inputs, scenario assumption controls, segmented filters, search, validation timing, and control layout. Covers the finance-specific cases that generic form advice gets wrong. Trigger on "form", "input", "date picker", "date range", "filter control", "validation", "assumption input", "scenario slider", "dropdown", "search box".
live-data-artifacts
Build artifacts that read live data from the viewer's claude.ai connectors - capability declaration, watchTool versus callTool, cache and freshness UI, error-code handling and retry policy, the observe-a-real-response rule, and graceful degradation to a baked snapshot when the capability is absent. Use when a page must show current data rather than a build-time snapshot. Trigger on "live data", "live artifact", "connector", "real-time dashboard", "pull from BigQuery in the page", "auto-refresh", "mcp capability", "make it live".
mermaid-in-artifacts
Use mermaid for diagrams inside artifacts - the one visualization that renders natively with no library, no CSP problem, and no hand-rolled SVG. Covers what it is genuinely good for (process flows, decision trees, close calendars, system maps, org and approval chains), what it is bad for (anything quantitative), and the syntax and theming that survive both light and dark. Trigger on "diagram", "flowchart", "process map", "sequence diagram", "gantt", "org chart", "decision tree", "mermaid", "how do I draw a process".
microcopy-and-states
The words in a data interface - labels, empty and error states, tooltips, disclosure and provenance lines, and the disclaimers a finance UI needs. Covers writing states that tell the reader what happened and what to do, and naming things so two people reading the same number agree on what it is. Trigger on "empty state", "error message", "what should this say", "label", "tooltip text", "wording", "microcopy", "disclaimer", "footnote", "caption".
motion-and-transitions
Use motion to aid comprehension in data interfaces - transitions that preserve object constancy when data changes, entry and update choreography, easing and duration choices, what should never animate, and honouring reduced-motion. Trigger on "animation", "transition", "animate the chart", "make it feel smooth", "easing", "reduced motion", "should this move", "loading animation".
stateful-artifacts
Build artifacts where the page itself is the record - trackers, checklists, sign-off boards, approval queues, polls, and collectors that remember what viewers do and share it across everyone who opens the page. Covers live-document sync regions, the classic publish model, conflict handling, read-only viewers, and local-only scratch UI. Trigger on "tracker", "checklist", "the page should remember", "shared state", "sign-off", "collect responses", "collaborative page", "save state in the artifact".
svg-charting
Build charts by hand in inline SVG when no charting library can be loaded - scales, nice ticks, axes, bars, lines, areas, stacking, waterfall layout, label collision avoidance, and responsive sizing. Ships chartkit.js, a tested dependency-free utility. Use whenever you need a chart inside a strict-CSP artifact, or anywhere a library is unavailable or not worth its bytes. Trigger on "chart without a library", "inline SVG chart", "no CDN", "D3 won't load", "hand-roll a chart", "axis ticks", "scale", "waterfall layout", "chartkit".
timeseries-finance
Time series for financial data - indexing and rebasing, log scale for compounding, TTM and run-rate versus point-in-time, seasonality handling, forecast boundaries, event annotation, period alignment, and the alternatives to dual axes. Trigger on "time series", "trend", "over time", "monthly chart", "YoY", "TTM", "run rate", "seasonality", "log scale", "two different scales", "index to 100".
ui-antipatterns
The tells that make a page read as templated, generated, or unconsidered - and what to do instead. Covers the AI-slop visual signature, dashboard cliches, false-precision and chartjunk, layout defaults nobody chose, and the specific ways financial UIs mislead. Read before shipping any page, and when a design is technically fine but feels generic. Trigger on "why does this look generic", "make it look designed", "AI slop", "looks templated", "generic design", "what's wrong with this UI", "chartjunk", "misleading chart".
uncertainty-charts
Visualizing forecast uncertainty, scenarios, and sensitivity - fan charts, scenario bands, tornado diagrams, Monte Carlo output, box plots, and the discipline of not implying more confidence than the model supports. Trigger on "forecast", "scenario", "uncertainty", "confidence interval", "fan chart", "sensitivity", "tornado", "Monte Carlo", "best case worst case", "range of outcomes", "what if".
variance-charts
Charts for actual versus plan, forecast, or prior - bullet charts, deviation bars, dumbbell and slope charts, variance bands, and small-multiple budget-vs-actual grids. Covers plotting the variance rather than two lines the reader must subtract, and encoding favourable-versus-unfavourable when direction and goodness disagree. Trigger on "actual vs budget", "BvA chart", "vs plan", "variance chart", "bullet chart", "target", "over/under", "compare to forecast", "gap to plan".
visual-hierarchy
Layout, spacing, grouping, and scanning order for data-dense pages - deciding what is most important and making the page say so, at densities a professional audience actually wants. Covers the rank-then-lay-out method, grouping by proximity, alignment discipline, the spacing scale, and layout patterns for dashboards, reports, and comparison views. Trigger on "layout", "hierarchy", "too cluttered", "where do I put this", "dashboard layout", "spacing", "grid", "it all looks the same", "make it scannable".
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.