IC Memo from Diligence Packet
Finance
Overview
Teaches an agent to turn a diligence risk register, financial narrative and comparables into a disciplined investment committee memo whose call is forced by the top open risk tier, whose valuation arithmetic is written out and divided back, and whose register rows are never silently reclassified.
Skill bundle · 2 files
SKILL.md
--- name: ic-memo-from-diligence-packet description: "Load when a user pastes a diligence risk register (with tier and status columns), a financial narrative and a comparables set for an M&A or investment target and asks for an investment committee memo, the committee write-up, or the diligence pulled into a recommendation. Returns a five-section IC memo (deal thesis, tiered risk table, valuation, recommendation, conditions) whose call is forced by the top open risk tier, whose valuation arithmetic is written out and divided back, and whose register rows keep their source status unless a contradicting sentence is quoted." --- # Reading a diligence packet the way a committee member does I have watched committees approve deals on a row everyone assumed was closed, and price deals on revenue that left with the seller. Both mistakes come from letting one artifact silently overrule another. The risk register, the financial narrative and the comp table were written by different people at different times; where they disagree, the memo says so in plain words and decides which fact governs. Every number in the memo is either stated in the packet or computed in view from stated numbers. ## A register status stands until another artifact overturns it The register's analyst saw the underlying documents; the memo writer sees three summaries. So I copy each row's tier and status exactly as written and treat `Resolved` as that analyst's finding, not as a claim waiting for proof. The finding text describes the exposure the analyst weighed: a pending application, a right a counterparty holds, a claim that was filed, a consent the deal needs. A row whose own text names the risk (who may cancel, who filed first, whose consent is needed, what claim was brought) is not reopened by that text, and the absence of a confirming document does not reopen it either. A `Resolved` row becomes **disputed** only when another artifact (the financial narrative, the comparables, an appended note) states a present fact the resolution cannot coexist with: a process it says is still running, a notice already served, a cure management cannot document. I quote that sentence and carry the row as open. Every other `Resolved` row stays resolved and does not move the call; I write that the source reports it resolved and list the record that confirms it among the routine conditions. In-Progress, Unresolved, blank and unknown statuses never count as clear, and a filing, a hearing date or a cure in progress on such a row is a timetable, not a closure. When a tier looks too low for what the row describes, I say why and leave the tier as given. ## The top tier decides the call Only the register's deal-breaker tier (`Must-Address`, `Critical` or its equivalent) can stop the deal; a register with no such row has no gate. When a deal-breaker row is open, blank or disputed, no form of Proceed is available, including Proceed with conditions. If the blocker can be cured before signing (a consent, a licence, an audit opinion, a settlement, a finished remediation), the call is **Defer**, and the deal comes back to the committee once the cure exists. The call is **No-Go** only when the packet already shows the cure has failed or cannot happen: a counterparty has refused, a regulator has denied, or a right the core business needs cannot be obtained by any document. A cure that is slow or uncertain still means Defer, and revenue that cannot transfer comes out of the base rather than ending the deal. With every deal-breaker row clear, the call is Proceed with conditions when open lower-tier rows need binding remedies, and Proceed when they do not. An asking price above my range is a counter-offer, not a reason to walk on its own. ## Value what the buyer will own The base comes before the multiple. Reported revenue is the base unless the packet says part of it will not stay with the buyer: one-time work, amounts collected for others when the peers report net, income that cannot move to a new owner, customers already lost. Revenue that is only at risk still transfers: customers who may not migrate, convert or renew, a contract that may lapse, a concentrated customer. It stays in the base; I say so in one sentence and let it move the band placement and the conditions. An add-back stays only if the packet shows it will not recur. Base, growth, margin and any share used together must cover the same period and scope: LTM with LTM, run-rate with run-rate, net with net, one segment with that segment. When two packet figures disagree (a stated blended margin and its segment margins, two revenue totals), I quote both, say which one the memo uses and why, and build no third figure from them. A segment share converts to the whole only when both shares cover the same period. When removing revenue leaves a base whose own growth or margin is not stated, I say so and do not credit it with the old growth. The range rests on one base: a revenue mix the comps do not report is a direction in words, never priced piece by piece. Each comparable is named with its disposition and the fact that decides it: an acquisition price typically carries a control premium and is shown apart from the trading band, a growth or margin profile far from the target prices a different business, a not-meaningful multiple drops on that metric only. Missing peer size or revenue mix is stated as a limit. Negative or unreliable EBITDA puts the valuation on revenue; positive EBITDA is at least a cross-check, and the memo says which metric sets the range and why. ## Arithmetic a committee can redo Every multiple is applied with its parts written out and rounded only at the end (`4.5x × $18.0M = (4 × $18.0M) + (0.5 × $18.0M) = $72.00M + $9.00M = $81.0M`), each endpoint is divided back by the base to confirm the multiple, and the ask is divided by the same base. I place the ask by its multiple against the band's two multiples (below, inside or above), not by a dollar gap. The target is compared with named peers; when it trails the included peers on growth, margin or retention, the band runs from the lowest included peer up to their median and not above it. The disposition a comparable gets in the table is the one the band paragraph uses. Equity value appears only when the packet states debt and cash: equity = EV − debt + cash. Those are the memo's only calculations besides the base equation and a one-line EBITDA cross-check on the included peers: no margin rebuilt from segments, no dollar gaps, no downside case, no second base, no implied peer margins, no averages. A risk that could shrink the base is named in words, not priced. ## The memo A title line `# Investment committee memo: <target>`, then five sections in this order with exactly these headings: `## 1. Deal thesis`, `## 2. Tiered risk table`, `## 3. Valuation`, `## 4. Recommendation`, `## 5. Conditions / next actions`. 1. **Deal thesis**: what we would buy and why, the quality of the earnings, the investment stance, and the primary risk named so that the same finding appears in the risk table. 2. **Tiered risk table**: columns `ID and finding | Source tier/status | Evidence about current closure | Effective state and consequence`, one row per source finding in the source's order, source IDs kept or analyst row numbers labelled as such, resolved rows retained. The closure column quotes the contradicting sentence or says that none is stated. 3. **Valuation**: metric and period, the base equation with its source, any reconciliation gap, the comp table with each disposition, the selected band and why, both endpoint equations with divide-back, the ask on the same base if one is stated, and the principal uncertainty. 4. **Recommendation**: the label first (`Proceed`, `Proceed with conditions`, `Defer` or `No-Go`), the rows that force it, the price view kept separate, and what is authorised now. 5. **Conditions / next actions**: numbered sentences, each tied to a named row or valuation gap, naming who acts, the record that confirms or cures it, the milestone (before signing, before close, after close) and what happens if it fails. Distinguish blocking cures from routine confirmation and post-close work. No invented dates, names, probabilities or thresholds. Follow any length bounds in the prompt; otherwise the memo runs as long as the packet warrants, with no caveat repeated. `references/worked-memo.md` is a complete memo for an invented target, followed by the Section 4 that two worse packets would have produced. Read it before you write, and only once. Only its shape and its checks carry over: do not reuse its sentences, and argue every row of the packet in front of you with that packet's own facts and figures. The reply is the memo alone, from its title on. ## Before you send - Does the call match what the risk table forces? If the recommendation could not be defended against the table row by row, redraft the recommendation, not the table. - Is the thesis's primary risk also a row in the risk table? If not, one of them is wrong. - Does the valuation name every comparable it left out of the band, with the fact that decided it? A silent exclusion is the commonest way a range goes wrong. - Is every condition tied to a named row or valuation gap? A floating condition means the table or the valuation is incomplete. ## What this skill does not do - It does not re-tier the risk register. When a tier looks wrong for what the row describes, say why and leave the tier as given. - It does not source comparables or rebuild the financial model. Work from what the packet provides. - It does not produce board-level or external communications (a press release, an offer letter, a letter of intent). The output is an internal committee memo. - It does not negotiate price. The range is diligence-derived, not a bid.
Platform-hosted file · no public repository.
AI
Scout Summary
Lab Performance
91Lab score
Won an open challenge, scored by independent validators.
Rating
No ratings yet
Details
Creator
sundae_bar Lab
Files
Bundle · 2 files
Lab Submissions
389
Security Analysis
Malware-free
Pass
File integrity
Pass
Reputable source
Pass
Installation
Install via CLI
Or download via curl