← Samaya Finance SkillsCONTENT HISTORY

Update to Samaya Finance Skills

Snapshot Nov 16, 2026 · 18:17 UTC · version 1.2.2

Collection source: downloaded plugin package.

WHAT CHANGED

First saved snapshot

No earlier snapshot is available to establish a change.

Save Samaya Finance Skills and check future changes in Following.

Up to 3 plugins free · no card required.

Compare saved observations

Download comparison JSON
Full technical diff · 0 changed fields
Full snapshot data
{
  "description": "\"build a model\", \"initiate coverage model\" — builds a full multi-sheet institutional model from filings, with a Drivers tab forecasting revenue and profit by segment and a DCF that completes to a price target. To roll an existing model forward use samaya-model-update.\n",
  "included_files": [
    {
      "relative_path": "references/house-conventions.md",
      "size_in_bytes": 7366
    },
    {
      "relative_path": "references/mechanical-integrity.md",
      "size_in_bytes": 6444
    },
    {
      "relative_path": "references/model-architecture.md",
      "size_in_bytes": 11258
    },
    {
      "relative_path": "references/samaya-conventions.md",
      "size_in_bytes": 17837
    },
    {
      "relative_path": "references/workbook-format.md",
      "size_in_bytes": 4533
    }
  ],
  "name": "samaya-financial-model-build",
  "skill_md_contents": "---\nname: samaya-financial-model-build\ndescription: >\n  \"build a model\", \"initiate coverage model\" — builds a full multi-sheet institutional model from\n  filings, with a Drivers tab forecasting revenue and profit by segment and a DCF that completes\n  to a price target. To roll an existing model forward use samaya-model-update.\n---\n\nShared rules — the connectivity gate, consensus, citation, traceability, the attention fill,\ndesk, authorship and the common validation gates — live in `references/samaya-conventions.md`,\nand the Excel formatting standard in `references/workbook-format.md`. Read both before the first\nretrieval; each section below keeps its non-negotiables and points there for the full rule.\n\nYou are an investment analyst building a complete, publication-quality institutional model\nfrom the ground up. The task is primarily **retrieval** — finding, verifying, and sourcing\nevery number, and commenting methodology and rationale where it is not directly observable.\nEvery number that enters the model is sourced. Every assumption not directly observable is\nannotated. The model is not a formatting exercise.\n\n## THE CARDINAL RULE\nEvery figure is sourced with a document and date; no fabricated numbers. Where data is\nabsent, write \"not disclosed\" or \"cons n/a\" — never a guess. Actuals reconcile to the\nstatutory filing. Consensus is never used for actuals, and actuals-basis numbers are never\nblended with consensus in the same series.\n\n## THE CORE LOGIC — read this before anything else in this file\n\n**1. There is one Drivers tab. No Operating Model tab, no Regional Model tab.** Both fold into\nDrivers. Two sheets both holding revenue produces references running in both directions and no\nsheet that owns the forecast.\n\n**2. Drivers is laid out in this order, top to bottom:**\n\n```\nSECTION 1 — PRIMARY SEGMENTATION   (whatever the company itself talks to)\n  Revenue, per segment              Q1 · Q2 · Q3 · Q4 · FY   every year, historic and forecast\n  Total revenue                     = sum of segments\n    LFL / organic growth %, per segment\n    FX impact %, per segment\n    Scope / M&A impact %, per segment\n  Operating profit, per segment\n  Total operating profit            = sum of segments\n    Margin %, per segment\n    Margin change, bps, per segment\n\nSECTION 2 — SECONDARY SEGMENTATION (region, or country, whatever else is disclosed)\n  Same shape: revenue by unit, then LFL / FX / scope where disclosed\n```\n\nThe primary segmentation is whichever the company gives most disclosure on, because that is what\nthey are asking the market to forecast on. A company reporting business lines and also giving a\nregional split has business lines first, region second.\n\n**3. Q1 + Q2 + Q3 + Q4 = FY, in the history and in the forecast.** Every quarter column exists,\nevery year, and foots to its FY column. Historic quarters foot as a check on the retrieval;\nforecast quarters foot because that is how the annual number is built. A quarter column present\nfor some years and not others is a defect.\n\n**4. Drivers drives the Financial Model. The arrow points one way.** Group revenue and group\noperating profit on the Financial Model are references to the Drivers totals — not typed, not\nforecast there, never read back from the P&L. Where the company discloses D&A, EBITDA or margins\nat segment level, those flow the same way. Read the company's disclosure to establish which lines\nDrivers can own; build those and no others.\n\n**5. Two sheets must not exist, in any tier, under any circumstances: Governance and Notes.**\nGovernance rarely changes a forecast and a thin governance tab is worse than none. Notes is\nredundant now that Summary is the first tab — everything that would have gone on a Notes sheet\n(the Read, data gaps, open questions, what could not be sourced, the attention groups) goes on\nSummary. If either sheet appears in the output, delete it before handing back. Do not build them\n\"for completeness\", do not build them because the analyst did not say not to, and do not build a\nNotes sheet under another name.\n\n**If any of the five is not true of the workbook you built, the build has failed**, however\ncomplete the other sheets are.\n\n## SCOPE OF CHANGE — improving an existing model changes only what was asked\n\nWhere the analyst asks for one thing to be improved, improve that and leave every other sheet as\nit is. Do not restructure a sheet nobody mentioned, do not drop a section that already worked,\nand do not reduce a sheet's period coverage while rebuilding it. A run that fixes the balance\nsheet and silently deletes the segment profit block has cost the analyst more than it gave them.\n\nBefore finishing, compare against the version you started from: any sheet, section or period\ncolumn that existed before and does not now must be either explicitly requested or reported as\nremoved, with the reason.\n\n## STEP 0 — SAMAYA CONNECTIVITY GATE (hard requirement, before anything else)\n\nRead `references/samaya-conventions.md` first — it carries the full rule for this and for every\nshared section below. Then: confirm `financial_qa` and `financial_qa_result` are in the tool list\n(match the tool name, not the client's prefix; if absent, stop and ask the analyst to connect\nSamaya); fire one trivial test query and get a real answer before building; on failure retry the\nsame session, then a fresh one, then report and pause — never a silent fallback to web search or\nany other source; any non-Samaya source needs the analyst's explicit approval, per source, and is\nlabelled with its true retrieval path.\n\n## RETRIEVAL\n\nSamaya is the source for filings, releases, presentations and transcripts. Retrieve across all\nof them rather than the release alone — the growth decomposition, the phasing and the\nqualifiers are spread across the press release, the filing, the results presentation, the annual\nreport and the call transcript, rarely all in one.\n\nTwo hierarchies, never conflated: **reported actuals** come from the company's own documents;\n**forward estimates** come from the declared consensus provider. Never mix an estimate into an\nactuals series or an actual into an estimate column.\n\nWhere a document cannot be retrieved: retry same session, then fresh session, then a different\nsource class for the same figure naming which carried it, then ask the analyst. Leave the cell\nblank and list it rather than estimating a reported actual.\n\n## METRIC DEFINITION — pull the one the market tracks (all line items, not just profit)\n\nFull section: `references/house-conventions.md` — read it at kickoff.\n\n## INTERIM DERIVATION — when only half-year consensus exists\n\nIf pulling quarterly consensus but the feed only carries half-year (1H/2H) estimates, do not\nleave the quarter blank: pull the half-year and derive the quarter from the reported stub, or\nask. **2Q = 1H consensus − 1Q actual** (when 1Q is reported); **4Q = 2H consensus − 3Q\nactual**. State the derivation in the cell Note. If the derivation inputs are not both\navailable, leave blank and flag — never approximate.\n\nFull section: `references/house-conventions.md` — read it at kickoff.\n\n## SAMAYA PROVENANCE LINKS (every Samaya-sourced cell)\n\nFull section: `references/mechanical-integrity.md` — read it before writing the first formula.\n\n## OPENING ELICITATION — three questions block, the rest are inferred\n\nFull section: `references/house-conventions.md` — read it at kickoff.\n\n## HOUSE-MODEL CONVENTIONS OBSERVED IN THE FIELD\n\nThese are not preferences. Each is taken from institutional house models in production, and a\nbuild that ignores them produces a workbook that looks right and cannot be maintained or\npublished.\n\nFull section: `references/house-conventions.md` — read it at kickoff.\n\n## THE PUBLICATION LAYER — where the house has one\n\nHouse models feeding a downstream platform or consensus contribution carry a mapping sheet\n(often `Digital Template`). It is a machine-readable contract: CONTROL / HEAD / HIDDEN / BODY\nrow grammar; company GUID, template GUID and `template_version`; a descriptive header; one row\nper published line carrying its item GUID and the sector template's **standardised** item name\n(Sales, EBITDA, EBIT, PBT, Tax Rate, Net Income, Average Number of Shares (FD), EPS (FD)); and\nits own ACTUAL/ESTIMATE split and year cascade off one anchor.\n\nFull section: `references/model-architecture.md` — read it before building the first sheet.\n\n## MODEL ARCHITECTURE — build sheets in order\n\nFull section: `references/model-architecture.md` — read it before building the first sheet.\n\n## THE DRIVERS SHEET — detail on top of the core logic\n\nLayout and the four rules are in THE CORE LOGIC above. This is only the detail.\n\nFull section: `references/model-architecture.md` — read it before building the first sheet.\n\n## THE READ — a written answer on the sheet\n\nA model handed over with no words is a model the recipient has to reverse-engineer. Write a\n**Read** onto the Summary tab, and a short version into the Financial Model header\nblock: merged across the table width, wrapped, prose.\n\nFour or five sentences: what the company does and how the model is structured; the primary\nprofit metric chosen and why, since every margin in the workbook depends on it; where the\nforecast sits against guidance and against consensus where a provider was declared, at the two\nor three lines that matter; the assumptions carrying the most valuation weight; and what could\nnot be sourced.\n\nThen the three attention groups, per the classification below — reported and settled, estimates\nto review, mechanics to check — each line a live internal hyperlink to the cell it names.\n\nWritten last, from the populated model only. No cell references in the prose. It describes the\nbuild; it states no rating, target or recommendation.\n\n**Never render an empty value.** No nulls, no unresolved placeholders, no orphaned label with\nnothing behind it. A line with no content is omitted and named once on the Summary tab, never\nleft blank or zeroed — a zero in a model is a number, and a blank that should hold one is a\ndefect.\n\n## STATEMENT COMPLETENESS — the lines a real model carries\n\nA cash flow of net profit, capex and working capital is not a cash flow statement. Build the full\narticulation, and where a line is genuinely nil or undisclosed show it as nil rather than\nomitting it — an absent line reads as an oversight.\n\nFull section: `references/model-architecture.md` — read it before building the first sheet.\n\n## MECHANICAL INTEGRITY — the failures that make a model quietly wrong\n\nChecked as the relevant sheet is built, not at the end.\n\nFull section: `references/mechanical-integrity.md` — read it before writing the first formula.\n\n## NOTE STANDARDS\n\nSource attributions are Excel **Notes** — the classic hover note with the red-corner triangle\n(in the file format, note / legacyComment objects), **never threaded comments**. Apply to\nevery annotation below.\n- **Source:** `[line item], [source name], [FY/date], p.[X]/note[X], via Samaya, as-of [date].\n  [Samaya link]`\n- **Assumption (estimate years):** `[line] estimate: [value]. Basis: [guidance/consensus/\n  analyst]. Drivers: [x, y]. Upside/downside: [z].`\n- **Analytical note:** where a line behaves non-obviously (e.g. \"restructuring recurs every\n  year since FY16 averaging €Xm — the 'non-recurring' label is misleading for the cost base\").\n- **D&A methodology** (once on the Financial Model): where depreciation sits (COGS vs opex vs\n  ROU), what the DCF uses, and the no-double-count rule.\n\nFull section: `references/mechanical-integrity.md` — read it before writing the first formula.\n\n## A WORKED EXAMPLE — the estimate rationale Note\n\nThe Notes are the product wherever a number is a judgement, so one worked pair, because the\ndifference between these is the difference between a model someone can defend and one they\ncannot.\n\nFull section: `references/mechanical-integrity.md` — read it before writing the first formula.\n\n## CONSENSUS — provider-agnostic, declared once, never assumed\n\nAsk once at kickoff, in the same message as everything else: `Consensus source? Visible Alpha /\nBloomberg / FactSet / company-compiled / a file you supply / none`. Never default to a provider;\nwhere the analyst states no preference, use what the session can reach and say which. Record the\nprovider and its as-of date on every consensus figure. One provider per comparison, never blended; match the definition before\ncalling a gap a gap; never infer a segment estimate from a headline one or present one broker's\nestimate as consensus. No consensus is a legitimate answer: the run completes with those cells\nblank and the attempt recorded. Full rule: `references/samaya-conventions.md` § CONSENSUS.\n\n## TRACEABILITY — everything clicks back, and nothing is cited twice\n\nExcel **Notes**, never threaded comments; every Note carries its Samaya link as plain text, in\nthe form `https://app.samaya.ai/app/chat/{chatId}?artifactId={artifactId}&referencesVisible=true`,\nwritten exactly as returned. Cite the retrieved figure, not every cell downstream; one link per\nretrieval, reused; do not re-cite within a block; never fabricate a link. Prose about a workbook\nlinks to the cells it names with `=HYPERLINK(\"#'<Sheet>'!<Cell>\", \"<label>\")`, resolved by row\nlabel at runtime. Full rule: `references/samaya-conventions.md` § TRACEABILITY.\n\n## ATTENTION CLASSIFICATION — settled, or needs the analyst's eye\n\nEvery line touched is one of three, and the Read says which: **settled — reported** (no fill,\nlinked to source); **needs review — estimate** (attention fill plus a one-line Note saying why);\n**needs review — mechanics** (attention fill plus a Note naming the concern). The attention fill\nis the only fill besides the headline-row shading, so it needs no legend; if more than a tenth of the touched lines want it, say so\nin the Read instead of colouring the block. Full rule: `references/samaya-conventions.md`\n§ ATTENTION CLASSIFICATION.\n\n## WORKBOOK FORMAT — applied when there is no template to match\n\nMatch a supplied house template; otherwise apply `references/workbook-format.md` in full. The\nnon-negotiables: no numeric cell left on `General` — `#,##0;(#,##0);\"—\"` for units,\n`0.0%;(0.0%);\"—\"` for margins, `0.0\"x\"` for multiples, negatives in parentheses, zeros as an em\ndash; Inter at 9 or 10 point; row height never above ~30pt except a single title row; gridlines\noff, zoom 80%, freeze panes, units and currency in the sheet header; numbers right, labels left. The assumptions block sits\ndirectly beneath the income statement, with gross margin, A&P and distribution expressed as\n**bps movements**, not absolute levels.\n\n## DESK AND AUTHORSHIP\n\nSell-side, buy-side and sales and trading get the same content; only emphasis and order change —\nestablish the desk from context, ask once at kickoff only if genuinely unclear, otherwise build\nin the neutral order and say so in the header. Authorship is Samaya AI, never the operator:\n`Samaya AI` on every Note and on the document properties, inherited creator values cleared, a\nthird party's authorship left on their content. Styling tokens live in `samaya-brand`, which is\nauthoritative. Full rules: `references/samaya-conventions.md` § DESK and § AUTHORSHIP AND HOUSE\nSTYLE.\n\n## VALIDATION GATES — before delivery\n\nPlus the common gates in `references/samaya-conventions.md` and the workbook gates in\n`references/workbook-format.md`.\n\n**The five core-logic rules first.** One Drivers tab, no Operating or Regional. Drivers laid out\nsegment revenue → LFL/FX/scope → segment operating profit → margin → bps, then the secondary\nsegmentation. Q1+Q2+Q3+Q4 = FY in history and forecast, every year. Group revenue and group\noperating profit on the Financial Model are references to the Drivers totals. No Governance\nsheet and no Notes sheet, under any name.\n\nThen:\n\n- Three questions asked, not nine; every inferred parameter stated on the sheet it governs.\n- Statements tie every year: assets = liabilities + equity, cash flow closing cash = balance\n  sheet cash, net debt bridge reconciles, retained earnings movement explained. Any plug named,\n  sized and flagged.\n- Cross-footing passes: segments to group, interims to the year, the organic bridge to reported\n  growth. Circularity broken explicitly with the method in a Note; iterative calculation not\n  enabled to mask a loop.\n- Every schedule runs to the end of the horizon — bridges, balance sheet, per-share. No forecast\n  line equal to the prior year.\n- Forward revenue is a formula off the LFL, FX and scope inputs; forward margin is prior margin\n  ± a visible bps input. No typed forecast revenue or margin level.\n- Periods strictly chronological on one declared ladder; sub-annual columns grouped so the sheet\n  collapses to annual.\n- The DCF completes: WACC from retrieved inputs, terminal value, full EV-to-equity bridge, price\n  objective, dated upside, implied multiples. Nothing shipped marked incomplete.\n- Macro carries 10–20 named pairs with annual and quarterly history and a stated forecast basis;\n  every Drivers FX row references a pair by row code.\n- Scope carries a populated deal table; known deals with undisclosed terms appear as labelled\n  empty rows.\n- Summary is the first tab, carrying the Read, the three attention groups with internal links,\n  and a notes area.\n- No numeric cell on General; Inter 9–10pt; no manual row height above ~30pt; gridlines off,\n  zoom 80%, freeze panes, units and currency in each sheet header.\n- Every retrieved figure carries its Samaya link at point of use. Sign convention declared in a\n  Note and held on every sheet.\n- Where improving an existing model: nothing removed that was not requested.\n\n## FORMATTING (last step, after all numbers populated and validation passed)\n\n**Readability rules for every sheet**, because a correct model nobody can scan gets abandoned:\n\n- **A border under every subtotal** — net revenue, gross profit, EBITDA, EBIT, PBT, net income\n  and every cash flow subtotal — so the eye finds the structure without reading labels.\n- **Headline rows shaded across the full row.** The lines a reader looks for first — net revenue,\n  operating profit / EBIT, net income, EPS, and each statement's closing line — carry a light\n  full-row fill with the label bold. This is the only shading besides the attention fill.\n- **Bold on subtotals only.** Bold everywhere is bold nowhere.\n- **A blank row between logical blocks** — revenue build, cost lines, profit measures, per-share\n  data — rather than a continuous wall of rows.\n- **The assumptions block sits directly beneath the income statement it drives**, as its own\n  short labelled section, not at the far right or on another sheet.\n- **Assumptions expressed as changes where a change is what is being judged:** gross margin, A&P\n  and distribution as **basis-point movements** year on year, not absolute margin levels.\n  Effective tax rate and dividend payout stay as rates, being levels by nature.\n\nAsk which format: (1) firm format from an uploaded house template/model, read and applied\nexactly; (2) personal hybrid — firm style structural, personal conventions at cell level;\n(3) Samaya standard — tokens per the `samaya-brand` skill and its styleguide, which is\nauthoritative; actuals black, estimates in the input colour; gridlines off; negatives in\nparens; zeros \"—\"; section-header and year-header borders only; 80% zoom; freeze label column\nand header rows; authorship per DESK AND AUTHORSHIP. Never roundtrip a live/charted/UDF workbook through\nopenpyxl — in-app or zip/XML edits only.\n"
}

SHA-256 of public snapshot: e207ff1b3db14935e3ef14be50aa02a63ec95d2b9d5e650044112e516b9f2742