Research skills. From prompt to note.

MCP gives agents governed tools. QJ Skills give them repeatable financial research workflows: evidence collection, route selection, output structure and compliance boundaries for buy-side and research teams.

See worked examples →MCP setup Explore the public skill pack ↗

Public skill pack

Eight workflows for the research process.

The public pack is a curated, installable subset of QJ workflows. It includes methodology and evidence standards without internal tenant assumptions, private tool maps or operating details.

Equity Deep Dive

Build a company note from profile, prices, fundamentals, metrics, estimates, events and ownership context.

Macro Regime Brief

Bring rates, the macro calendar, benchmarks and commodities into one market-regime brief.

Earnings Preview / Recap

Prepare for earnings, compare results with consensus and track surprises, estimate changes and follow-up questions.

Institutional Flow

Review 13F ownership, holder concentration and reported flows with filing delays and coverage caveats attached.

Valuation Workbench

Organize comparables, ratios, estimates and scenarios around explicit assumptions and sensitivities.

Thesis Tracker

Maintain a living thesis as events, evidence, open questions and coverage gaps change over time.

Risk Bias Review

Test a thesis against the bear case, missing evidence, coverage limits and potential research biases.

Portfolio Monitor

Track event risk, factor context, exposure changes and watch items for a portfolio review.

How it fits

Tools need a research method.

API routes provide the data. MCP exposes approved tools. Skills define how an agent gathers evidence, handles gaps and turns results into a reviewable research note.

Governed tools

Use tenant-scoped MCP tools with provider metadata and audit context.

Repeatable method

Follow a playbook for tool selection, evidence collection, gap handling and output structure.

Evidence-backed output

Keep facts, interpretation, sources, caveats and open questions visible for review.

From prompt to research packet

One request. A defined output.

Build a deep dive for MSFT before the investment meeting. Include fundamentals, valuation, recent events, ownership context, risks and open questions.

Evidence plan: Profile, adjusted prices, statements, ratios, estimates, filings, insider activity and ownership.
Agent behavior: Use governed MCP tools, retain provider metadata, separate facts from interpretation and flag missing coverage.
Research packet: Company snapshot, fundamentals and valuation, events and filings, ownership and flow, risks and open questions.

Shared output contract

Every note records the tools used and gaps in the available evidence.

sections:
  - company_snapshot
  - fundamentals_and_valuation
  - events_and_filings
  - ownership_and_flow
  - risks_and_open_questions

required:
  tools_used: true
  coverage_gaps: true
  no_investment_advice: true

Open methodology

Read it. Adapt it. Version it.

The public repository contains Markdown skills, examples, shared evidence standards, output contracts and compliance boundaries. Review the instructions and adapt them to your team's research process.

Open qj-agent-skills on GitHub ↗

Six worked examples. Full research notes.

A skill tells the agent which routes to call, in what order, which caveats to attach and how to write the note — so facts, interpretation and open questions never blur. Open-source files you can read, change and version like code.

Open an example to see its prompt, research note, evidence and coverage gaps.

Equity deep dive

Company research note built from profile, prices, fundamentals, metrics, estimates, events and ownership context.

Prompt

Use qj-equity-deep-dive on Micron ahead of the 30 September print.

Answer

Exhibit 1 · Positioning against the peer set

FINRA consolidated short interest, settlement 31 Aug 2026.

Exhibit 2 · The numbers behind the call

P/E trailing · EV/EBITDA	22.7× · 16.5×
Revenue y/y	+345.7%
ROIC · ROE	44.0% · 70.5%
Net debt / EBITDA	−0.27× (net cash)
Income quality	1.02
Capex / revenue	28.0% at 2.8× depreciation
Inventory days · cash cycle	126 · 198 days
Consensus, 30 Sep	$31.27 EPS on $50.59bn

Interpretation

Cheapest multiple in large-cap semis on the fastest growth, funded from net cash rather than leverage. Income quality of 1.02 says the earnings are arriving as cash — the distinguishing feature against peers below 1.0. The 198-day cash conversion cycle is the memory cycle showing through the balance sheet, and it is the channel through which a demand pause hurts before it reaches the P&L.

Open questions

No ASP or bit-shipment data in the catalog — the pricing assumption underneath +345.7% cannot be verified here.
Inventory is reported in aggregate; the DRAM/NAND split is not on these routes.
Guidance is unpublished; all forward figures are consensus, not company.

Coverage 3 institutions · settlement to 31 Aug, fundamentals TTMTools equity.fundamentals.get_key_metrics_ttm · equity.ratios.get_pe_ratio_ttm · equity.ratios.get_quarterly_revenue_growth_yoy · equity.shorting.get_short_interest · equity.calendar.get_earnings_calendar

Macro regime brief

Rates, macro calendar, benchmarks and commodity context summarised into a market-regime read.

Prompt

Give me this morning's qj-macro-regime-brief, and include the funding plumbing.

Answer

Exhibit 1 · Volatility, rebased

All three rebased to 100 at 18 Aug 2026. SKEW above the base while VIX and VVIX sit below it.

Exhibit 2 · The regime, by institution

Reading	Level	Source
CPI y/y · unemployment · claims	3.35% · 4.1% · 196k	bls · fred
2s10s · 2-year move in Sep	+25bp · +37bp	fmp
Total public debt outstanding	$40.09tn · +$354bn in 6 weeks	treasury
SOFR, 17 Sep	3.85% · +23bp in a day	fred
Specs, 10Y futures	−812.5k contracts	cftc
VIX · SKEW	14.81 · 148.1	cboe
Fear & Greed · BTC funding	29 · +8.04% p.a.	cnnf · ccxt
October hike priced	54.5% · +17pt in a week	polymarket

Interpretation

Late-cycle reflation with an unhedged equity market and a front end under supply pressure. Prices are re-accelerating into a tightening labour market while the Treasury added $354bn of debt in six weeks; SOFR printing 3.85% on 17 September is that supply arriving in the funding market. Spot volatility is below where August began while SKEW is above it — convexity is bid, level is not. Sentiment reads fear while positioning shows nothing defensive. The asymmetry is not in the hike; it is in how little is hedged against it.

Open questions & data gaps

BLS published no CPI observation for October 2025. Carried through, not interpolated; any 12-month calculation crossing it is affected.
Fear & Greed returned a flat 7-day series (volatility 0.0) — treat the level, not the trend.
COT is settled to 15 Sep; the 37bp front-end move continued after that date.
The SOFR jump is a single observation; quarter-end effects are not separated here.

Coverage 8 institutions · 7 domains · oldest input 15 Sep, newest liveTools macro.us.get_cpi · macro.series.get_jobless_claims · macro.series.get_secured_overnight_financing_rate · fixedincome.reference.get_debt_to_penny · benchmarks.treasury.get_treasury_rates · macro.positioning.get_net_positioning · derivatives.vol.get_vix_data · derivatives.vol.get_vvix_data · derivatives.vol.get_skew_index_data · sentiment.fear_greed.get_summary · crypto.funding.get_funding_rate · predictions.markets.search_markets

Earnings preview & recap

Pre-earnings setup, consensus context, surprise recap, estimate drift and post-event watch items.

Prompt

Run qj-earnings-preview on the Micron print. What has consensus done into it?

Answer

Exhibit 1 · Target drift, by publication window

Average published target per window; sample size in brackets. Dashed line is spot.

Exhibit 2 · The sample

Window	Published	Average target	vs spot
Last 30 days	2	$1,300.00	+28.0%
Last quarter	24	$1,557.08	+53.3%
Last 12 months	102	$793.28	−21.9%
All time	260	$376.30	−63.0%

Interpretation

Estimate drift into the print is negative: the 30-day average sits 16.5% below the 90-day average. At the same time the stock trades 21.8% below even the reduced average. Either the street is walking down a target it will not defend, or the market is discounting a cycle turn the street has not yet written down.

Open questions

n = 2. The 30-day average rests on two publications — a signal about coverage thinness as much as about price.
Publishers are aggregator-level, not a mapping to underwriting banks.
Post-event watch items cannot be set until guidance is published.

Coverage 2 institutions · targets to 19 Sep, calendar to 5 OctTools equity.calendar.get_earnings_calendar · equity.estimates.get_price_target_summary · equity.market.get_quote

Institutional flow

13F ownership, holder concentration and delayed institutional-flow context — with the filing-lag caveat attached automatically.

Prompt

qj-institutional-flow on NVDA — who owns it and is the crowd still adding?

Answer

Filing lag — attached automatically

Holdings are Q2 2026 13F filings, position-dated 30 June 2026, filed through 14 August. As of today they are 81 days stale, and the quarter-end price of $200.09 is 10.0% below the current $222.27. Nothing below describes a current position.

Exhibit 1 · Who moved, and which way

100 filers returned by this route; breadth is added ÷ (added + cut).

Exhibit 2 · Concentration and the counterweight

Institutional shares	14.47bn across 100 filers
Top-10 concentration	56.3% · BlackRock alone 8.03%
Net change in the cohort	+829M shares
Short interest	298.3M · 2.14 days to cover
Long-to-short ratio	48 : 1

Interpretation

Concentrated but not one-sided. The net add is driven by the index complex rather than conviction managers, and breadth at 51% is a coin flip. With a 48:1 long-to-short ratio there is almost no positioning counterweight, so flow-driven downside would have little to cover into.

Open questions

13F covers long US equity positions only — no shorts, no derivatives, no non-US holders.
The cohort is the top 100 filers by this route, not the full filing universe.
Q3 filings are due mid-November; the conclusion has a known expiry date.

Coverage 2 institutions · positions 30 Jun · short interest 31 AugTools regulatory.ownership.get_institutional_holders · equity.shorting.get_short_interest · equity.market.get_quote

Valuation workbench

Comps, ratios, estimates and scenario valuation with assumptions written down and a sensitivity table.

Prompt

qj-valuation-workbench on Microsoft at 56× free cash flow — what has to happen for that to make sense?

Answer

Exhibit 1 · The gap the valuation has to close

Free cash flow yield versus the 18 Sep ten-year and August CPI. The valuation carries a −318bp spread to the risk-free rate.

Exhibit 2 · Scenario, assumptions stated

Path	What must change	Verdict
A · capex normalises	FCF ≈ 100% of operating cash flow, i.e. capex → 0	arithmetically impossible
B · cash flow grows	operating cash flow +174% at an unchanged capex ratio	possible, demanding
C · multiple compresses	neither of the above; the gap closes through price	the residual

Interpretation

The valuation does not rest on capex normalising — the arithmetic rules that out. It rests on operating cash flow roughly tripling, or on accepting a negative spread to the risk-free rate for an extended period. That is a defensible position; it is a different one from “capex is temporary”.

Open questions

Enterprise value held constant; no terminal-value or WACC assumption is made here.
Capex is a single line — the maintenance-versus-growth split is not on these routes.
The risk-free rate is the 18 Sep close, not an average.

Coverage 3 institutions · fundamentals TTM, rates 18 Sep, CPI to AugTools equity.fundamentals.get_key_metrics_ttm · benchmarks.treasury.get_treasury_rates · macro.series.get_cpi

Risk & bias review

Bear case and bias review that challenges a prior thesis against evidence, coverage and risk signals.

Prompt

Run qj-risk-bias-review against my long NVDA thesis: dominant position, 105% growth, 59% ROIC, street sees 54% upside.

Answer

Exhibit 1 · The metric the thesis omits

Income quality = operating cash flow ÷ net income, trailing twelve months. Below 1.0 means reported earnings exceed cash generated.

Evidence the thesis does not address

Income quality	0.70 — cash below earnings
Inventory days · cash cycle	150 · 154 days
Days to cover	2.14 — no short base to squeeze
13F breadth, Q2	51% — half the cohort reduced

Bias review

Three of the four supports in the thesis are restatements of one fact — growth. The consensus target is cited as confirmation, but a 53.8% average upside across 22 publications in 30 days is a crowded view, not an independent one. No falsifier is stated anywhere in the thesis.

What would falsify it

Income quality staying below 1.0 for a second consecutive period.
Inventory days extending beyond 150 without a revenue acceleration.
13F breadth dropping below 45% in the November filings while concentration rises.

Coverage 3 institutions · fundamentals TTM, filings 30 Jun, settlement 31 AugTools equity.fundamentals.get_key_metrics_ttm · equity.estimates.get_price_target_summary · equity.shorting.get_short_interest · regulatory.ownership.get_institutional_holders

Why it matters

Without a method, every prompt re-improvises the analysis and every analyst gets a different note. With a method, the same question produces the same shape of answer — today, next quarter, for anyone on the desk.

What a skill guarantees

Every output separates Facts (what the data said) from Interpretation (what the agent concluded) and Open questions (what it could not establish), and ends with Coverage and Tools used. Caveats such as the 13F filing lag are attached automatically.

How to use one

Connect the MCP endpoint, clone the skill pack, and name the skill in your prompt — “Use qj-equity-deep-dive on Micron.” The skill picks the routes, the gateway resolves the institutions, the envelope records who answered.