Bank tier classification

Skill jefrnc/quant-llm-skills/skills/bank-tier-classification

Skills for quant research with LLMs that don't fall for the traps nobody talks about. Lookahead bias, ATM detection, survivorship, transaction-cost realism — 10 skills + 11 reproducible evals showing 8 of them measurably improve Haiku output. Claude Code & Cursor compatible, MIT.

Install

npx -y skills add jefrnc/quant-llm-skills --skill bank-tier-classification

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when an SEC filing names a placement agent, underwriter, or sales agent (e.g., Goldman Sachs, H.C. Wainwright, Maxim, Aegis, Roth, B. Riley, Cantor, Jefferies). Classifies the firm into a 4-tier framework that materially changes the dilution risk profile of the offering.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

7.2 KB, ~1.7k tokens by cl100k_base, as published. Nobody here has run it

Bank Tier Classification

The same filing (424B5, 8-K) means very different things depending on who is selling the shares. A registered direct led by Goldman Sachs is a different animal than an ATM run by a tier-4 specialist. Most quant pipelines ignore the agent identity entirely; this skill restores it as a primary signal.

Core principle

The placement agent is a leading indicator of the offering's structure, aggression, and post-deal price behavior.

Tier 1 firms protect their reputation and underwrite institutional demand. Tier 4 firms specialize in placing offerings that tier 1 firms will not touch — by definition, those are the harder, more dilutive, more structured deals.

Same form, different agent = different trade.

The four tiers

Tier 1 — Bulge bracket

Global investment banks underwriting at scale. Their participation implies institutional demand and a "blessed" issuer.

Goldman Sachs (and GS&Co.)
Morgan Stanley
J.P. Morgan / J.P. Morgan Securities
Bank of America Securities (BofA / Merrill)
Citigroup Global Markets / Citi
Barclays Capital
Deutsche Bank Securities
UBS Investment Bank
Credit Suisse (legacy / now UBS)
Wells Fargo Securities

Implication when seen on a small-cap filing: rare. If present, the issuer is a real institutional name or has a strategic reason for the relationship. Dilution risk: low to moderate, deal is typically clean.

Tier 2 — Mid-tier full-service

Quality investment banks with research coverage and institutional distribution. Do legitimate growth-capital raises.

Jefferies
TD Cowen / Cowen
Stifel Nicolaus / KBW (Keefe Bruyette Woods)
Piper Sandler
Raymond James
William Blair
Cantor Fitzgerald
Needham & Company
Truist Securities
Mizuho Securities
Nomura Securities
BMO Capital Markets

Implication: legitimate growth capital. Often firm-commitment underwritten secondaries, not ATMs. Dilution risk: moderate, predictable.

Tier 3 — Boutique / sector-specialist

Reputable smaller banks with niche specialties (biotech, energy, tech). Mix of clean and aggressive deals depending on the issuer.

Roth Capital Partners / Roth MKM
Oppenheimer & Co.
B. Riley Securities / B. Riley FBR
Lake Street Capital Markets
Craig-Hallum Capital Group
Canaccord Genuity
Wedbush Securities
Northland Capital Markets
D.A. Davidson
Stephens Inc.
Compass Point
JMP Securities (now Citizens)

Implication: mixed bag — read the structure (warrants, discount, ATM language) carefully. The agent identity alone is not decisive at this tier. Dilution risk: moderate to high depending on deal terms.

Tier 4 — Specialized small-cap placement agents

Firms that specialize in placing offerings for issuers that cannot attract Tier 1–3 underwriting. Heavy concentration in ATMs, registered directs with warrants, ELOCs, and PIPEs. High-volume, high-frequency small-cap deal flow.

H.C. Wainwright & Co. (the dominant small-cap placement agent by deal count; runs many ATMs)
Maxim Group
Aegis Capital
A.G.P. / Alliance Global Partners
EF Hutton / Kingswood (legacy entity, now separate)
ThinkEquity
Dawson James Securities
Joseph Gunnar & Co.
Spartan Capital Securities
Univest Securities
Boustead Securities
Brookline Capital Markets
Laidlaw & Company
WallachBeth Capital

Implication: when one of these agents appears on a 424B5 or sales agreement, the base rate for: (a) attached warrants, (b) deep discount to last trade, (c) ATM that sells aggressively into intraday strength, (d) repeat dilution within 30–90 days, is materially elevated. Dilution risk: HIGH. Treat the issuer as having a continuous selling pressure floor until the offering is exhausted or terminated.

Decision rules

"Is this offering institutional or retail-flow?" Tier 1–2 = institutional book. Tier 4 = retail / quant / market-maker absorption with frequent shorting against the cross.
"How fast will this dilute?" Tier 4 ATMs sell every green tick. Tier 2 underwritten secondaries are one-shot. The agent tier is the strongest predictor of cadence.
"Should I go long after this offering closes?" Tier 1–2 underwritten secondaries often see post-deal stabilization (the underwriter has incentive to support the price). Tier 4 ATMs have the opposite incentive — keep selling.
"Multiple agents listed — which tier wins?" Use the LOWEST (most aggressive) tier present. A "Tier 2 + Tier 4" joint placement is dominated by the tier-4 dynamics.

Important caveats

The list is not exhaustive. Many smaller agents exist; classify based on observable behavior (ATM frequency, warrant attachment rate, client universe) when in doubt.
Tier is not a moral judgment. Tier 4 firms perform a legitimate market function placing offerings that would otherwise fail. The classification predicts deal behavior, not firm quality.
Firms can move tiers over time. B. Riley's profile, for example, has shifted across cycles. Re-validate against recent deal flow rather than relying on a frozen list.
Same firm, different deal. A Tier 4 agent doing a one-shot registered direct (firm commitment, no warrants, single investor) is much less aggressive than the same agent on an open ATM. Always read the structure alongside the tier.
Avoid defamation. When writing public-facing content, frame as "specialized in" / "frequently associated with" rather than absolute toxicity claims about specific firms.

Workflow when an agent name appears

Identify the agent in the 424B cover or "Plan of Distribution".
Place into Tier 1–4 using the lists above.
Inspect the structure (ATM vs. registered direct vs. underwritten secondary; warrants attached; discount to market).
Combine tier + structure for the final risk read:
- Tier 1–2 + underwritten secondary = clean
- Tier 3 + registered direct = read the warrants
- Tier 4 + ATM = continuous dilution overhang
- Tier 4 + ELOC = deeper-discount slow bleed
Cross-reference with atm-detection if the structure is an ATM, and lookahead-safety for any historical timing claims.

Phrases that should trigger this skill

Any of the firm names above appearing in a filing context
"placement agent" / "lead underwriter" / "sales agent"
"who's the bank on this deal"
"is Wainwright running this" / "is Maxim doing this"
"tier 4" / "toxic shop" / "bulge bracket"
"Plan of Distribution"

What this skill is NOT

This is not a firm-rating service. It is a heuristic that maps observed firm names to expected deal behavior based on documented patterns of small-cap deal flow. Always combine with structure (atm-detection) and timing (lookahead-safety) for a complete read.