From quant-llm-skills
Use when determining whether a company has an active At-The-Market (ATM) offering, controlled equity offering, or equity line (ELOC). Defines the multi-signal inference required to distinguish ACTIVE dilution from dormant shelf capacity, and disambiguates ATM from ELOC.
How this skill is triggered — by the user, by Claude, or both
Slash command
/quant-llm-skills:atm-detectionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
ATM offerings are the dominant slow-bleed dilution mechanism in small caps.
ATM offerings are the dominant slow-bleed dilution mechanism in small caps. Detecting them requires inference from multiple co-occurring signals — no single filing or phrase is sufficient. Most quant pipelines either miss them entirely (looking only for 8-K item 3.02) or flag every S-3 as ATM (false positives that drown the signal).
An ATM is ACTIVE only when all three are true:
Missing any one of these = NOT an active ATM (could be dormant shelf, could be a registered direct, could be a private placement).
| Signal | Where it lives | What to look for |
|---|---|---|
| Effective shelf | S-3 / S-3ASR / F-3 with effective status | effectiveDate is set |
| Placement agent retained | 424B5 / 424B4 cover or "Plan of Distribution" section | Named agent (one or more) |
| ATM authorization language | 424B prospectus supplement OR underlying sales agreement (often filed as Ex-1.1 or Ex-10.x to an 8-K) | See phrase list below |
| Active ATM (not just authorized) | Subsequent 10-Q "Stockholders' Equity" or "Subsequent Events" note | Discloses shares sold under ATM during period, OR confirms agreement remains in effect |
The first two confirm STRUCTURE. The third confirms TYPE (ATM vs. registered direct). The fourth confirms ACTIVITY (in-use vs. dormant).
In the prospectus supplement or sales agreement:
These mechanisms all use a shelf and dribble shares into the market, but they are NOT the same and have different toxicity profiles.
| Mechanism | Counterparty | Pricing | Discretion | Typical agreement language |
|---|---|---|---|---|
| ATM | Placement agent (bank) selling into open market | Market price | Issuer-controlled volume | "Sales Agreement" / "at-the-market" |
| ELOC | Specific investor (e.g., Lincoln Park, Yorkville, B. Riley) | Discount to recent VWAP | Issuer-initiated puts | "Common Stock Purchase Agreement" / "Equity Line" / "Standby Equity Purchase Agreement" |
| Registered Direct | One-time investor(s) | Negotiated, typically discount | Single-shot | "Securities Purchase Agreement" + 424B5 covering only that takedown |
ELOCs are typically MORE toxic per share than ATMs (forced discount, counterparty incentivized to short into puts) but often slower per day. A registered direct is one-and-done — confirm via 8-K item 1.01 + 3.02 within days.
False positives (saying "ATM active" when it isn't):
False negatives (missing an active ATM):
lookahead-safety. Whatever you conclude is true
only as of the LATEST filing in your set. An ATM established last
week was not active last month.Active ATM + low cash runway (<6 months) in last 10-Q + recent strength (green day, gap up) = highest-probability dilution window. Issuers with active ATMs sell aggressively into strength because that is when they get the best price.
This is not a parser. It does not extract numbers from filings. It
provides the inference rules needed to interpret the parsed data
correctly — i.e., to decide whether a set of filings constitutes
an active ATM. Combine with sec-filing-types for form semantics,
lookahead-safety for time semantics, and (when available)
bank-tier-classification for placement-agent toxicity grading.
Searches MemPalace before answering questions about past work, people, projects, or prior decisions. Returns verbatim stored content instead of guessing from model memory.
Guides Payload CMS config (payload.config.ts), collections, fields, hooks, access control, APIs. Debugs validation errors, security, relationships, queries, transactions, hook behavior.
Implements vector databases with Pinecone, Weaviate, Qdrant, Milvus, pgvector for semantic search, RAG, recommendations, and similarity systems. Optimizes embeddings, indexing, and hybrid search.
npx claudepluginhub jefrnc/quant-llm-skills --plugin quant-llm-skills