knowledge-base
Captures precedents, multiples, lessons, and reusable templates from a completed deal into a structured, searchable form so each deal makes the next one faster.
SKILL.mdpacks/investment-banking/skills/knowledge-base/SKILL.md
GitHubKnowledge Base Agent
When to use
Use this agent when you want each deal to make the next one faster: at the end of a transaction to capture what was learned before the team disperses, or when building a reusable library of precedents and multiples across many deals. It is for turning finished work into a searchable asset rather than letting it evaporate. Run it while memories are fresh, ideally within days of close.
What it does
Produces a structured knowledge capture of precedents, multiples, lessons, and templates, organized and tagged so the next team can find the relevant prior work quickly.
Method
Capture the deal fact sheet.
- Record sector, transaction type, size, parties (anonymized where required), timeline, and outcome.
- Keep it to a scannable block so a future reader can triage relevance in seconds.
Record the precedent transaction data.
- Log entry and exit multiples, EV/EBITDA and EV/revenue at announcement, premium paid, and financing structure.
- This is what makes the deal reusable as a comparable in future valuation work.
Log the valuation approach and key assumptions.
- Note the methods used and the assumptions that drove the answer, with a pointer to the final model.
- Future teams reuse the method, so record why, not just what.
Capture lessons learned in two buckets.
- What worked and should be repeated, and what went wrong and should be avoided.
- Phrase each as an actionable takeaway, not a description of events.
Extract reusable templates and artifacts.
- Preserve the working group list, the diligence request list, the process letter, and any bespoke analyses worth keeping.
- Point to where each artifact lives rather than duplicating it.
Tag every entry with consistent metadata.
- Sector, deal type, year, and theme, using a controlled vocabulary so entries surface in future searches.
Note the contacts and advisers involved.
- Record their roles so relationships are preserved as institutional memory.
File the capture under the agreed taxonomy.
- Store it in the shared repository and cross-link it to related prior deals.
Inputs
- The closed deal's key facts, terms, and outcome
- Final valuation model and the multiples achieved
- The premium paid and the financing structure used
- Team reflections on what worked and what did not
- The working documents worth preserving as templates
- The shared repository location and its tagging taxonomy
Output format
Claude returns a structured knowledge entry with these sections:
- Deal fact sheet: sector, type, size, parties, timeline, outcome.
- Precedent transaction data: multiples, premium, and structure, described in prose.
- Valuation approach: methods, key assumptions, and a pointer to the model.
- Lessons learned: split into repeat and avoid, each an actionable takeaway.
- Templates and artifacts: named, with pointers to where they live.
- Metadata tags: sector, deal type, year, and theme. It reads as a filing-ready record that slots into a searchable library alongside other entries.
Example
Entry: Northwind Logistics, sale of cold-chain division, sell-side advisory, enterprise value near four hundred million, closed this year, competitive auction. Precedent data: EV/EBITDA of 9.5x at announcement, EV/revenue of 1.4x, control premium of about thirty percent over the standalone plan, financed with senior debt and a sponsor equity check. Valuation approach: DCF with a WACC of nine percent cross-checked against precedent transactions; key assumption was mid-single-digit volume growth. Lessons, repeat: running a tight two-round process kept tension and lifted the final bid. Lessons, avoid: the tax structuring started too late and cost a week at signing, so engage tax counsel at kickoff next time. Templates preserved: the auction process letter and the diligence request list, both pointed to in the shared drive. Tags: logistics, carve-out, auction, sell-side, current year.