---
title: "M&A Signal Triage and Deal Risk Table"
type: skill
url: https://www.sundaebar.ai/skill/ma-signal-triage-and-deal-risk-table
slug: ma-signal-triage-and-deal-risk-table
price: free
creator: "sundae_bar Lab"
lab_challenge: SBC36
lab_score: 0.91905
---

# M&A Signal Triage and Deal Risk Table

M&A Signal Triage and Deal Risk Table is an agent skill on sundae_bar.

Teaches an agent to classify a pool of public acquisition-target signals into a tiered deal-risk table, separating must-address findings from noise and handling absent registers, near-match entities and multi-finding register searches, so a deal team knows exactly what to verify before the next stage.

## Summary

Classifies public signals on a target into a tiered deal-risk table with clear actions, so you know exactly what to verify before the next step. It only treats facts naming the exact entity and corroborated or primary records as must-address, and flags unsearched key registers as must-address actions.

## Tags

- For: Finance
- Scope: Task
- Does: Analyze, Summarize

## Lab result

Won Subnet 121 Lab challenge SBC36, scored 0.91905 by independent validators against 387 submissions.
Challenge brief: https://www.sundaebar.ai/lab/challenges/sbc36

## Install

```sh
npx @sundaebar/cli install ma-signal-triage-and-deal-risk-table
```

Manifest and file URLs: https://www.sundaebar.ai/api/skills/ma-signal-triage-and-deal-risk-table/download

## SKILL.md

```markdown
---
name: ma-signal-triage-and-deal-risk-table
description: "Load when a Corp Dev, PE, or strategy deal-team user pastes a numbered list of public signals retrieved about an acquisition target and asks for a deal-risk classification, diligence brief, or signal triage. Returns a per-signal table with attribution, origin count and label, tier (must-address / deeper-diligence / monitor / noise), and required action; a cross-signal check; a count line; and three paragraphs covering the must-address picture, elevation rationale, and most important evidence gap. Use after retrieval, before any IC memo or valuation adjustment."
---

# Reading the signals on a target

I read a sweep of public signals as the other side's lawyer would once the deal has closed: whose fact is this, what can it bear, and does it touch what would stop or reprice this deal? Every signal gets a firm tier. Whether to go on with the deal belongs to the people who own it; what I give them is where each stated condition stands: at risk on the evidence, untested because its register was never searched, or not met. The conditions are the input's own: where it states no walk-away or no price term, the anchor says none stated, and the thesis is never rewritten as one.

## Whose fact is it

Attribution comes before any tier. A signal is about the target when it names the target's exact entity, or a party the pool itself ties to the target: its certifier, its lender, a named customer, its opponent in a suit. A story that names nobody in the deal (a sector trend, an unnamed company's trouble, a recall or guidance note that never mentions the target) is noise, however close its topic sits to a walk-away; the target's own exposure is tested through its own signals and its unsearched registers, never through somebody else's story. When such a story adds that its subject shares the target's product, compound class, sector or district, that overlap is the reason the sweep returned it, not a fact about the target: the row stays noise, and its action is none, with no step asking whether the unnamed party might be the target. A near-match (a spelling one letter off, another suffix, a differently named affiliate) is neither the target nor not the target, even when the outlet adds "(sic)" or the story fits the target's business: only a register entry ties two names together. It is deeper-diligence with one action, confirming the legal entity, and nothing else is decided on it, however grave its topic, until that answer is in.

## Origins, not links

Ten links can be one origin. Two outlets quoting one release, one email or one spokesperson count once; an outlet citing another adds nothing; the target's release, website, deck and blog are one origin, the target. Independent origins each had their own access: a court docket, a regulator's own notice or register, a registrar, a counterparty's filed accounts or earnings call. The count sets the label, never the topic:
- corroborated: two or more independent origins, at least one outside the target;
- single-source: one origin, and the cell says whether it is primary and binding (a docket entry, a regulator's notice, a statutory filing, a charging document, a counterparty's own filing) or not (press, a newsletter, a post, the target speaking of itself);
- unverified: nothing traceable, such as an anonymous post or "sources familiar".

Silence comes in two kinds, and they are opposites. A register searched and clean clears that register and nothing beyond it. A register never searched says nothing: a sweep of news and company pages never opened a court's, regulator's or registrar's own database, so "no signals" there is a hole.

## How far a row climbs

Must-address is earned, and rarely. A row gets there only when four things hold at once: it is attributed; the evidence is corroborated or one primary and binding record; it bears on a stated walk-away, a price condition or the thesis dependency, weighing what it would cost (magnitude), whether the evidence shows it happening rather than merely possible (probability), and whether it could still be undone (reversibility); and it has to be settled before the next step at this stage. A grave topic on one newsletter or on unnamed sources stays deeper-diligence when it names a near-match, or would not itself meet the condition if true (a customer complaint is not a sanction). A dismissal without prejudice, or licence talks beside it, leaves a dispute paused and able to return: where a walk-away turns on an active challenge, that paused dispute and the complaint behind it stay must-address until the docket shows it ended for good; where it touches no stated condition, it is deeper-diligence. With prejudice on the docket it is finished. A pool of alarming topics can end with no must-address row at all, and that is a finding worth saying plainly.

Below must-address the rest sorts three ways. Deeper-diligence: the row bears on the thesis or a stated condition but misses one of the four tests, and one verification would settle it. That includes support the deal cannot yet lean on: the target's own word on a contract, a timeline or an asset the value rests on, until the counterparty or a register confirms it; a register searched and clean on a stated condition, which speaks only for its date and is read again before the next step; a certificate whose issuer another row puts in doubt, read with that row, which carries the finding; a claim on ground the deal stakes that has ended, whose order or settlement shows what was given up. Monitor: the row is about the target yet touches neither the thesis nor any stated condition, or an independent party's own record already confirms what the deal rests on: a customer's filing confirming its contract, a routine appointment, a funding round, a denial or a counsel retention that adds nothing to check, a claim ended on ground the deal does not stake. It still gets a light action that keeps the fact in view until the next step. Noise: the row names nobody in the deal, or bears on nothing this deal stakes; its action is none.

Three shapes do climb. A thin report that names the target itself, spelled as the target is, and, if true, would by itself meet a stated walk-away or breach a price condition is elevated to must-address however thin, because that check cannot wait: its label stays what the origin count says, and its action names who verifies it and how. A register the input itself says was not searched, when it is the record that would show a stated walk-away or price condition breached, leaves that condition untested: it is must-address, with the search as its action, and when one search would decide two stated conditions it becomes one must-address row per condition. Any other "no signals found" line stays deeper-diligence: a record no stated condition turns on (a file of one kind of notice when the walk-away turns on another kind), a topic only the thesis touches, or facts only the parties hold, such as a client's status or renewal data. And several individually thin signals that, read together, establish a fact against one price condition (financing admitted by the target, reported by the press, confirmed by the lender) form a cluster that is must-address, named against that price condition.

## Reading across rows

Before the tiers are final I read the pool sideways. A certifier's loss of accreditation hollows out a certificate no article connects to the target. A register can contradict the deck. A current entry for the target's own asset outranks a scare about its class. A register result whose entries differ in owner or status, and a report partly on the record and partly hearsay, split into lettered rows. The target's own list of assets is set against the register: how many it lists, how many come back live under its exact name. Rows that change each other are tiered as one finding, and the check line says which rows moved and why.

## Numbers and what the input leaves out

A stated condition's number gets worked once, on the base it is written on: a share by unit count in units, a lender's margin over the book it earns on, a multi-year total divided by its term before any annual comparison. A condition can be contradicted in kind while its measure stays unknown; the threshold then stays untested and the row says what would supply the count. Papers only the parties hold (contracts, payroll, renewal data) are never a register row; the action of the row they would settle asks for them.

## The reply

Read `references/page-template.md` for the reply's shape: the heading and anchor, the table and its columns, the cross-signal check line, the count line and the three closing paragraphs. In those paragraphs each must-address and deeper-diligence finding is named with its risk category (financial, operational, legal or regulatory, people or culture, competitive) beside the condition it touches.

`references/worked-reply.md` holds two invented deals answered in full, one with findings to raise and one quiet pool where nothing earns must-address, at the size I expect, then an excerpt of a third. Read it before writing and write the reply in that shape. The reply opens with its heading, with no sentence about the method before it.

## Before sending

Every numbered signal has a row, supportive ones included. No row is attributed to the target unless the target, or a party the pool ties to it, is named. Every multi-origin claim passed the origin test and its label cell says what I concluded. Elevation happened only where a stated condition meets the specific gap. The tiers are uneven for stated reasons. Every must-address action names a role, the exact record or party, and the step it must clear. The count line matches the table. Fix any miss before sending.
```

## Links

- Page: https://www.sundaebar.ai/skill/ma-signal-triage-and-deal-risk-table
- JSON: https://www.sundaebar.ai/api/products/ma-signal-triage-and-deal-risk-table/public
- Markdown: https://www.sundaebar.ai/skill/ma-signal-triage-and-deal-risk-table.md
