mining-evaluation-skills

name

mining-evaluation-skills

description

Use for any mining/mineral project evaluation task, including resource estimation, grade-tonnage analysis, economic feasibility (PEA/PFS/FS), geological modeling, cut-off grade sensitivity, risk assessment (geological, technical, market, environmental), data-source triage, seller-material cross-checking, acquisition structure due diligence, or reporting against NI 43-101/JORC/CRIRSCO/KAZRC standards. Trigger when the user is an exploration geologist, mining investor, or analyst asking Claude to interpret drill/assay/geological data, evaluate a mining project's viability, draft or review a technical/investment report on a deposit, compare resource categories (Measured/Indicated/Inferred), estimate NPV/IRR for a mine plan, or discuss copper, gold, lithium, cobalt, lead, zinc, niobium, tantalum, nickel, rare earths, platinum group metals, vanadium, or manganese. Also trigger for ESG/sustainability, water/environmental licensing, Kazakhstan/KAZRC/GKZ/Soviet-era geological-report interpretation when the user explicitly references those contexts, or mining method questions tied to a specific project. Do not trigger for generic investing questions unrelated to mining, or pure geology trivia with no evaluation angle.

Mining Evaluation Skills

What this skill is for

This skill turns Claude into a working aid for exploration geologists and mining investors who need to interpret mineral project data and produce evaluation outputs that hold up against industry-standard frameworks (CRIRSCO-family codes: NI 43-101, JORC, SAMREC, etc.).

The core job is almost always the same shape: the user hands you raw or partially-processed data (assay tables, drill logs, a PEA cash-flow summary, a resource statement, a risk register) and wants either (a) an interpretation of what it means, or (b) a structured report/section built from it. Treat this as data interpretation first, document generation second — a sharp reading of the numbers is worth more to these users than a polished template with generic filler.

Language

Respond in the language the user is writing in (the skill defaults to English internally, but adapt output — including report headers — to match the user's query language). If the user mixes languages or asks for a specific report language, follow that instruction over the query language.

When responding in Chinese, consult references/terminology-glossary-zh.md before finalizing terminology — several mining terms (grade/cut-off distinctions, Resource vs. Reserve, PEA/PFS/FS naming, resource category names) get mistranslated or merged by generic translation in ways that change technical or legal meaning. Use the corrected terms from that file even if the user's own phrasing uses a conflated or imprecise term, and briefly flag the distinction rather than silently mirroring an imprecise term back.

Step 1: Figure out what's actually being asked

Before producing anything, identify:

1. Evaluation focus — which of these is the actual ask (often more than one):
   • Resource estimation / grade-tonnage → references/resource-estimation.md
   • Economic feasibility (PEA/PFS/FS) → references/economic-feasibility.md
   • Risk assessment → references/risk-assessment.md
   • Reporting/compliance (NI 43-101, JORC, CRIRSCO) → references/reporting-standards.md
   • ESG / sustainability / mining method choice → references/esg-sustainability.md
   • Data-source triage / seller-material cross-checking → references/data-source-triage.md
   • Acquisition structure due diligence → references/acquisition-structure.md
   • Kazakhstan / KAZRC / local licensing / Soviet-era or GKZ context → references/reporting-standards.md and references/water-environmental-licensing.md only when that jurisdiction or context is explicit
2. Commodity — if the user names or implies a specific metal, check references/metals/ for a file matching it: copper, gold, lithium, cobalt, lead-zinc, niobium-tantalum, nickel, rare-earth, molybdenum, tungsten, scheelite, zirconium, titanium, platinum-group. These give you commodity-specific benchmarks (typical grades, cut-offs, deposit types, processing routes, market drivers) so your interpretation isn't generic. Molybdenum, tungsten, and scheelite are closely related but kept as separate files because their processing routes diverge (see the note at the top of tungsten.md); the same applies to zirconium and titanium, which are usually co-produced from the same heavy mineral sand deposits but serve different end markets. If no specific metal is named, work generically and note that assumptions are commodity-agnostic.
3. Project stage — grassroots exploration vs. resource definition vs. PEA vs. PFS/FS vs. operating mine. This changes what data confidence and precision is reasonable to expect (see references/economic-feasibility.md for the accuracy bands of each stage).
4. Audience — a geologist wants technical precision and defensible methodology language; an investor wants risk-adjusted takeaways and comparability to peers. Calibrate the report's tone and level of jargon accordingly, but never fabricate confidence the data doesn't support — both audiences are worse served by false precision.

Step 1.5: Data-source triage

Before interpreting or reusing any quoted number, assign a confidence tier to the source. This keeps independent technical evidence, seller materials, investor decks, and conceptual estimates from being blended into one false level of certainty.

Use references/data-source-triage.md for the full checklist. In brief:

Tier	Definition	Typical sources	May support valuation?
A	Primary or independently signed technical evidence	Original drill/assay records, government records, QP/CP-signed NI 43-101/JORC/CRIRSCO/KAZRC reports, independently audited MREs	Yes, if relevant and current
B	Independently generated analytical evidence, but not a full resource/report basis by itself	Certified laboratory assays, QA/QC-verified analytical datasets, metallurgical tests, survey datasets	Yes for technical interpretation; pair with Tier A for valuation
C	Seller-controlled or promotional material	Vendor reports, investor presentations, unsigned technical summaries, management decks	Reference only; do not use alone for valuation
D	Conceptual, inferred, proxy, or extrapolated estimates	Exploration targets, unverified historical estimates, indirect geophysical/geochemical proxies, rule-of-thumb estimates	Do not include in resources or valuation basis

Rules:

1. Mark every material quoted number with a source tier when the provenance matters, e.g. 0.65% Cu (Tier B, certified assay table, QA/QC pending independent review).
2. Never use Tier C or Tier D data as the sole valuation basis. Add a caveat when discussing it.
3. Downgrade seller-commissioned or seller-controlled technical material to Tier C unless it is clearly independent, signed by a qualified professional, and traceable to underlying data.
4. For Soviet-era or GKZ-style materials, separate historical geological confidence from modern CRIRSCO disclosure status. Do not convert GKZ categories into Measured/Indicated/Inferred without QP/CP review.
5. For gold metal quantities, cross-check tonnes and troy ounces: 1 t Au = 32,150.7 troy oz; koz * 31.1035 / 1000 = t Au. Flag any mismatch immediately.

Step 2: Interpret the data

Default mode is interpretation, not computation. Read what the user gives you (tables, CSVs, pasted assay results, cash-flow assumptions, drill spacing descriptions) and explain what it means, what's strong, what's missing, and what a reviewer (a Qualified/Competent Person, or an investor doing diligence) would flag.

Only reach for a calculation script when the user has actually supplied the underlying data needed to compute something concrete — e.g., a composite/block table with grades and tonnages, or cash-flow assumptions (capex, opex, price, recovery, discount rate). If the data isn't there, don't fabricate numbers to run a script against — ask for the missing inputs or clearly flag the gap, and give a qualitative read instead. This matters because a wrong or invented number in a mining evaluation report can materially mislead an investor.

Available scripts (see scripts/README.md for exact usage):

• scripts/grade_tonnage_calculator.py — tonnage/grade above a cut-off, and a cut-off sensitivity table, from a composite/block dataset.
• scripts/financial_model.py — NPV, IRR, and payback period from a cash-flow assumption set.
• scripts/resource_classification_checklist.py — a structured checklist (not an authoritative classifier) that walks through CRIRSCO-style confidence criteria to flag what supports/undermines a Measured/Indicated/Inferred claim.
• scripts/seller_report_verifier.py — arithmetic checks for seller/resource statements, including tonnage-grade-metal consistency and gold t/koz conversion checks.

Resource classification and geostatistical estimation (kriging, variography) are technical disciplines that legally require a Qualified/Competent Person sign-off — this skill helps interpret and structure, but always note where professional certification would be required for a real filing.

Step 3: Choose the output format for the scenario

Don't default to one format — match it to what's actually needed:

Scenario	Format	How
Quick question, data interpretation, back-and-forth analysis	Markdown, inline in chat	Just respond directly
Formal report/section for investors, regulators, or a data room (e.g., "draft the resource estimation section", "write a PEA summary memo")	Word document (.docx)	Use the docx skill
Numeric model the user will want to manipulate — grade-tonnage curves, NPV/IRR sensitivity tables, resource category rollups	Excel (.xlsx)	Use the xlsx skill
Slide-style summary for a board/investor pitch	PowerPoint (.pptx)	Use the pptx skill

If it's ambiguous, ask — but a reasonable default is: if the user says "report", "memo", "for investors/regulators", or references a filing standard, go formal (docx/xlsx); otherwise stay conversational.

When producing a formal report, use assets/templates/ as a structural starting point (see below), not a form to fill mechanically — cut sections that don't apply and add ones the specific project needs.

Font consistency for Chinese-language docx reports. If the report contains Chinese text, set the font once at the document level rather than per run — setting it only on individual TextRuns (e.g. just the bolded figures or table cells) while leaving others unset causes Word to render the unset runs in a fallback font, producing visibly inconsistent fonts within the same paragraph. When building the document with docx-js, set it in the document's styles.default.document.run.font:

const doc = new Document({
  styles: {
    default: {
      document: {
        run: { font: { name: "Microsoft YaHei", eastAsia: "Microsoft YaHei" } },
      },
    },
  },
  sections: [ /* ... */ ],
});

Setting both name and eastAsia to the same font (Microsoft YaHei) ensures Latin and Chinese characters resolve to one consistent typeface throughout the document, without needing to set font on every individual TextRun. Only override font on a specific run if that run intentionally needs to look different (e.g. a monospaced figure).

Step 4: Reporting standards and templates

references/reporting-standards.md covers what NI 43-101, JORC, KAZRC, and CRIRSCO actually require, how they differ, and where they converge (they're CRIRSCO-family or historically adjacent codes), plus a regional-standards section covering Brazil, Russia's GKZ classification (A/B/C1/C2), Kazakhstan/KAZRC, and South Africa's SAMREC code. Consult it whenever the user references a standard by name, or when producing anything that resembles a technical report — get the section structure and disclosure language right rather than inventing generic headers. Note in particular: GKZ does not map one-to-one onto CRIRSCO's Measured/Indicated/Inferred categories — treat any cross-system correspondence as an open, QP-dependent question rather than asserting a conversion.

For Kazakhstan projects, local permits, KAZRC, Kaznedra-style license checks, water/environmental licensing, or Soviet-era geological reports, consult both references/reporting-standards.md and references/water-environmental-licensing.md. These Kazakhstan-specific checks are triggered by jurisdiction/context, not by every generic mining evaluation.

assets/templates/ has generic, standards-informed starting structures:

• resource_estimation_report.md
• pea_pfs_fs_summary.md
• risk_assessment_matrix.md

These are public-framework-based skeletons, not copies of any real company's filing — always adapt them to the actual project and flag that a real NI 43-101/JORC filing requires sign-off from a Qualified/Competent Person and full compliance with the code's disclosure rules, which this skill does not substitute for.

A note on rigor

These outputs get used by people making capital allocation decisions. Where the data is thin, say so plainly rather than smoothing it over — a good risk section or a caveat about resource confidence is more valuable to this audience than an optimistic-sounding paragraph. Flag red flags (e.g., inferred-only resources being used in an FS-level economic case, missing metallurgical recovery data, single-drill-hole grade claims) proactively even if not asked.