Fiscal Skills

A skill is an out-of-the-box prompt that harnesses the power of the Fiscal MCP for your research. Each one teaches your AI agent how to run a specific analyst task leveraging Fiscal.ai data: which endpoints to call, how to keep every number source-linked, and what a finished deliverable looks like. Invoke one by name ($fiscal-comp-set) or just ask naturally ("compare Amazon to its peers") and the agent picks the right one.

Everything a skill produces is built on as-filed data: filing-backed numbers carry click-through audit links that open the source PDF at the exact page in the Fiscal terminal.

Pick your skill

Onboarding & demo

trial-tour — the guided demo. A free-plan showcase that runs entirely on the trial ticker allowlist with a hard 8-call budget: three companies across three exchanges, side-by-side profile and valuation cards with terminal links, plus top headlines. Names a ticker outside the allowlist? It suggests the closest substitute. Try: "Show me what Fiscal.ai can do."

Company data & financials

financials-pull — statements fast, any depth. Standardized income statement, balance sheet, or cash flow laid out over any period range (annual, quarterly, LTM) with every figure audit-linked to its filing page. Apples-to-apples across companies by default; as-reported filed line items on request. Minutes, not an hour: this is the skill for history without a model. Try: "Pull Microsoft's income statement for the last 8 years with sources."

financial-model — the full three-statement model. Every line item in the company's reporting template across the last 10 fiscal years plus 10 quarters, with common-size, growth, and indexed-trend analytics, statement-level filing citations, and a forecast scaffold wired to driver assumptions: enter revenue growth, margin ratios, and a tax rate, and the projection populates itself. Template-aware: a bank, insurer, REIT, or utility gets its own statement structure. Default delivered as a spreadsheet. Try: "Build a financial model for Uber."

segments-and-kpis — what drives the business. Segment revenue and profit splits plus the operating KPIs that actually move the stock, examples being backlog and deliveries for an aerospace name, GOV and members for a marketplace, MAUs for a platform, same-store sales for a retailer. The data is rollup-aware so nothing double-counts, with discontinued segments shown as historical only. Try: "Break down Boeing by segment, including backlog and deliveries."

ratio-trend — multiples vs. own history. Daily time-series of any multiple Fiscal.ai computes (P/E, EV/EBITDA, P/S, P/B, FCF yield, etc) so "is it cheap versus its own history" gets an actual chart and percentile, not a feeling. Try: "Show Costco's P/E over the last five years vs. its average."

price-and-capitalization — price, splits, cap. Price history (back decades for mature names), period returns from 1-day to 5-year, 52-week range, split history, shares-outstanding trajectory (the buyback read), and the capitalization stack (market cap, TEV, net debt, and yields).

adjusted-earnings-quality — GAAP vs non-GAAP. Reconciles reported results against company-adjusted figures (adjusted EPS, EBITDA, FCF) period by period with absolute and percentage gaps, audit links into the reconciliation pages, and a read on where the divergence localizes (one charge quarter vs. chronic adjusting). Try: "How big is the gap between Meta's GAAP and adjusted earnings?"

Research & analysis

company-snapshot — one-page tear sheet. Identity, latest financials, valuation ratios, price action, and headlines on one screen. The fastest answer to "how is this company doing?" Try: "Give me a snapshot of Shopify."

investment-research — the 2,000-word deep dive. A buy-side-style note in eight sections: history and business model, competitive landscape, moat, full financial and KPI analysis, valuation vs. history and peers, price action and the bull/bear debate, latest earnings, management and insiders, recent news. Fiscal data first for every number; web sources (always attributed with URL and date) only for what Fiscal doesn't carry — transcripts, consensus, proxy compensation. Try: "Write a research note on DoorDash."

comp-set — side-by-side peer comps. Two to six companies on one table (valuation, profitability, growth, leverage) normalized to one currency, with warnings when reporting templates make a metric non-comparable. Names you choose, or peers discovered from the company's industry classification. Try: "Compare Amazon to its retail peers."

industry-report — full sector breakdown. Discovers the peer universe from a seed company or industry name, deep-dives the top 20 by size, and writes the report: how the economics evolved over the available history, the landscape table today, who's positioned how, the dynamics that decide winners, and where the directional signals point. Revenue share labeled as a proxy, since Fiscal carries no market-share data. Try: "What's the story in passenger airlines?"

valuation — reverse DCF, multiples, IRR. Picks the method to match the question: reverse DCF for "what's priced in," forward DCF for intrinsic value, relative multiples, dividend discount, or an expected-return/IRR model. Every forward assumption is yours, explicit, and sensitivity-tested. Composable with financial-model drivers. Try: "What growth is priced into Nvidia at today's multiple?"

capital-allocation — 10-yr sources & uses. Where the cash came from and where it went across five buckets (organic capex, M&A, buybacks, dividends, debt) over 5–10 years, with execution-quality lenses: buyback average price vs. the market, dividend coverage and growth, capex intensity, and the ROIC trajectory that says whether any of it worked. Try: "Has Apple's management been disciplined with cash over the last decade?"

Screening & monitoring

screener — filter the 12,300-name universe. Two-stage screen: descriptive filters (sector, industry, country, exchange, template) narrow the universe in one cheap pass, then metric thresholds (valuation, growth, margins, returns, leverage, yield) rank the survivors. Transparent about bounds: the metric stage caps at 50 names and the skill tells you exactly what was fetched vs. matched. Try: "Find Canadian non-financials with >4% FCF yield, clean balance sheets, and growing revenue that return capital to shareholders."

watchlist-monitor — the what-changed dashboard. Your coverage list (tickers or an industry), grouped by attention level: who reported, what moved, what filed, what hit the news, each flagged name with the why and a source link. Ships two ways from one data pass: a chat summary and a self-contained HTML dashboard with sortable return columns and price sparklines, built for re-opening every morning. Up to 25 names per run. Try: "Monitor my watchlist: CSU, TPZ, CROX, UBER, FRU, CVE."

news-and-events — ranked material news. Company or market-wide news scored by importance (1 = major event to 5 = routine) and classified by event type: earnings, guidance, M&A, buyback, regulatory, legal, executive, restructuring, and more. Leads with what changed and why it matters. Windows: 31 days per company, 7 days market-wide. Try: "Any news on Air Canada this month?"

ownership-activity — insiders & 13F. The conviction read: open-market insider buys and sells (properly separated from grants, exercises, and tax withholding), cluster detection, the current insider roster and stakes, institutional holders with quarter-over-quarter accumulation, and reverse lookup of any fund's portfolio. Every transaction linked to its SEC filing. US coverage only (Form 4 / 13F). Try: "Are insiders buying Broadcom? Who's been adding institutionally?"

filing-finder — 10-Ks/10-Qs with links. The filing index for any covered company: annual and interim reports, 8-Ks, earnings releases, and prospectuses. These all come with filing dates, fiscal periods, public source links, and amendments flagged as potentially authoritative over the originals. Can render a specific page on request. Try: "Find RTX's latest 10-K and recent 8-Ks."

sum-of-the-parts — what the parts are worth. Values a multi-segment company one division at a time: a peer multiple on each segment (EV/EBIT where it discloses segment profit, EV/Sales where only revenue), the unallocated corporate line reconciled back to the whole, then a bridge through net debt, minorities, and preferred to a per-share value versus today's price. Every multiple is your assumption, stated explicitly and sensitivity-tested on the segment that swings the answer. Needs real segment disclosure; single-segment names route to valuation. Try: "Are Amazon's parts worth more than the whole?"

credit-and-solvency — leverage, coverage, liquidity. A balance-sheet risk read across the leverage stack (net debt/EBITDA, debt/equity, debt-to-capitalization), coverage (interest, EBITDA-minus-capex, cash-flow-to-debt), and the current, quick, and cash liquidity ratios, with the multi-year trajectory that says whether the company is deleveraging or stretching. Leads with the directionally clean gauges and recasts the frame for banks, insurers, and other special templates instead of reporting a meaningless net-debt/EBITDA. No maturity schedule or agency rating in the data, so refinancing and covenant headroom stay qualitative and point to the filings. Try: "Can AT&T comfortably service its debt?"

quality-scores — the academic quality and distress scores. The named scoring models that boil a company's financials down to a few signal numbers telling you how financially sound it is, how close to distress, and whether the accounting looks manipulated. It computes the Piotroski F-Score (nine pass/fail tests of profitability, leverage, and efficiency, scored 0 to 9), the Altman Z-Score (a bankruptcy-distress gauge), the Beneish M-Score (a statistical flag for earnings manipulation), the Sloan accruals ratio (how much of profit is cash versus accrual), the Graham number (a deep-value price ceiling), and the Magic Formula ingredients (earnings yield and return on capital), each shown with its inputs and a plain-English zone rather than a bare number. Template-aware: models built for an industrial balance sheet are gated out for banks and insurers instead of being forced onto numbers they don't fit, and because these are heuristics, not verdicts, a distress or manipulation flag overrides a strong fundamentals score. Try: "Score Delta on quality and distress risk."

earnings-reaction — how it trades on earnings. Reconstructs the stock's earnings-day behavior from reported actuals and daily prices: the per-quarter reaction across the last 8 to 12 prints, the post-earnings drift, the EPS and revenue track record, and how large and how consistent the report-day move tends to be. Surprise is framed against the company's own prior and year-ago quarters, and the price move stands as the market's verdict (the reactions are raw, not market-adjusted). No forward calendar; for who reports next, see watchlist-monitor. Try: "How does Chevron usually trade on earnings?"

Good to know

• Plan features. News-powered skills need the news feature; ownership-activity needs the ownership feature. Skills detect the gap and report "unavailable on this plan" rather than pretending there's no data.
• Insider/13F coverage is US-ticker only, for now (SEC Section 16 / 13F). Non-US names report this as a coverage limit, never as "no buying."