Deal Workbench Review
This skill teaches a systematic method for completing M&A deal review tasks against the deal workbench API. The core pattern is: read the template first, gather all deal data from the workbench, compare draft terms against playbook rules or policy thresholds, compute dollar amounts from the headline purchase price, and produce JSON that conforms exactly to the provided answer template.
Key Principle
The task always provides an answer_template.json payload. This is your contract for output shape. Read it first and treat every field, enum, and constraint in it as non-negotiable. The workbench API provides the facts; the template defines what shape those facts must take.
Connecting to the Workbench
The task always provides the base URL as <TASK_ENV_BASE_URL> (or similar placeholder). Substitute it everywhere. The read-only SQL endpoint, when available, is POST <BASE>/api/query with body {"token": "deal-workbench-readonly", "sql": "<single SELECT or WITH statement>"}.
The primary API surface for a deal DEAL_ID consists of these endpoints:
| Endpoint |
Purpose |
GET <BASE>/workspace |
Workspace overview |
GET <BASE>/api/deals/<DEAL_ID> |
Deal record (headline price, parties, structure) |
GET <BASE>/api/deals/<DEAL_ID>/terms |
Current draft terms |
GET <BASE>/api/deals/<DEAL_ID>/risk-estimates |
Risk estimates for the deal |
GET <BASE>/api/deals/<DEAL_ID>/employees |
Employee data |
GET <BASE>/api/deals/<DEAL_ID>/consents |
Required third-party consents |
GET <BASE>/api/deals/<DEAL_ID>/regulatory |
Regulatory filings and status |
GET <BASE>/api/deals/<DEAL_ID>/benchmarks |
Market benchmarks for terms |
GET <BASE>/api/deals/<DEAL_ID>/documents |
Deal documents |
GET <BASE>/api/deals/<DEAL_ID>/notes |
Negotiation notes |
GET <BASE>/api/deals/<DEAL_ID>/cap-table |
Capitalization table (stock deals) |
GET <BASE>/api/deals/<DEAL_ID>/material-contracts |
Material contracts |
GET <BASE>/api/deals/<DEAL_ID>/diligence-findings |
Diligence findings |
GET <BASE>/api/playbooks |
Available playbooks |
GET <BASE>/api/playbooks/<PLAYBOOK_ID>/rules |
Playbook rules for a side |
GET <BASE>/api/policies |
Available policies |
GET <BASE>/api/policies/<POLICY_ID>/thresholds |
Policy thresholds |
GET <BASE>/api/search |
Search across the workbench |
The older non-/api prefixed routes (/deals/<id>, /playbooks, /policies) return the same data and can be used interchangeably with the /api variants. Use whichever the task prompt references.
Systematic Workflow
Follow these steps in order. Complete each step before moving to the next.
Step 1: Read the Answer Template
Read input/payloads/answer_template.json in full. Map every field:
- Top-level required fields and their types
- Allowed enum values for every constrained field (risk_rating, issue_status, recommended_action, etc.)
- Stable identifier lists (possible_issue_ids, stable_redline_ids, etc.) — these define the universe of expected outputs
- Numeric constraints (integer dollars, percent points to N decimal places, integer months)
- Array element shapes — the exact fields each object must have
The template is your schema. Every null in the template marks a place you must fill from workbench data.
Step 2: Gather the Deal Record
Fetch GET <BASE>/api/deals/<DEAL_ID>. Extract:
- The deal ID (confirm it matches the task)
- The headline purchase price (this is your calculation basis for all dollar amounts unless a source explicitly states a different basis)
- The parties (target, buyer, seller names)
- The deal structure (stock purchase, asset purchase, merger)
- The client side implied by the task context (buyer-side or seller-side counsel)
Step 3: Gather the Terms
Fetch GET <BASE>/api/deals/<DEAL_ID>/terms. These are the current draft terms the counterparty has proposed. Map each term to its stable ID. Pay attention to:
- Indemnity caps, baskets, and survival periods
- Escrow amount, duration, and release triggers
- Financing conditions and reverse break fees
- Restrictive covenants (non-compete, non-solicit)
- Employee provisions (service credit, PTO, retention)
- Tax allocation and transfer tax splits
- Governing law and forum
- Consent conditions and material contract treatment
- Closing conditions and outside dates
- Materiality scrapes, knowledge qualifiers, and sandbagging
Step 4: Gather the Rules or Policies
Depending on the task type:
- Playbook comparison tasks: Fetch
GET <BASE>/api/playbooks/<PLAYBOOK_ID>/rules. The playbook defines preferred and fallback positions for each term category. Compare each draft term against the playbook's preferred, fallback, and prohibited positions.
- Policy threshold tasks: Fetch
GET <BASE>/api/policies/<POLICY_ID>/thresholds. The policy defines approval thresholds that draft terms must not exceed without committee escalation.
Step 5: Gather Supporting Records
Fetch all endpoints the task references. Common supporting data:
- Employees (
/api/deals/<ID>/employees): headcount, PTO liabilities, service credit needs, WARN risks
- Consents (
/api/deals/<ID>/consents): which third-party consents are required, which are notice-only
- Regulatory (
/api/deals/<ID>/regulatory): HSR filing status, thresholds, effort standards
- Benchmarks (
/api/deals/<ID>/benchmarks): market data for term comparison
- Risk estimates (
/api/deals/<ID>/risk-estimates): quantified risk estimates
- Material contracts (
/api/deals/<ID>/material-contracts): contracts requiring consent or notice
- Cap table (
/api/deals/<ID>/cap-table): for stock deals, holder breakdowns
- Diligence findings (
/api/deals/<ID>/diligence-findings): issues from due diligence
- Documents (
/api/deals/<ID>/documents): relevant deal documents
- Notes (
/api/deals/<ID>/notes): negotiation context
Use the read-only SQL endpoint for cross-table checks when needed — send POST to /api/query with {"token": "deal-workbench-readonly", "sql": "..."}.
Step 6: Perform the Analysis
This step depends on the task type. Common analyses:
Playbook comparison (buyer or seller APA/SPA review):
- For each draft term, look up the corresponding playbook rule
- Classify the issue status:
draft_exceeds_playbook: draft is more favorable to counterparty than playbook's fallback (e.g., higher cap, longer survival)
draft_below_playbook: draft is less protective than playbook requires (e.g., lower escrow, missing fee)
missing_required_term: a term the playbook requires is absent from the draft
in_policy / out_of_policy: for policy threshold tasks
- Determine risk rating (HIGH/MEDIUM/LOW) based on dollar exposure and closing impact
- Determine recommended action (revise/add/delete/accept)
- Calculate dollar deltas:
draft_amount - fallback_amount or fallback_amount - draft_amount
Committee escalation (policy threshold):
- For each draft term, compare against the policy threshold
- Exclude terms that are in-policy or stale
- For out-of-policy terms, compute the delta between draft and threshold
- Provide benchmark context and quantified exposure
Closing package / transition review:
- Identify missing provisions the client side requires
- Classify each gap and provide the required position
- Produce redlines showing what must change
Step 7: Compute Dollar Amounts
All dollar amounts derive from the deal's headline purchase price unless a specific source (term, finding, risk estimate) explicitly states a different basis.
amount = headline_price × (percent / 100), rounded to integer dollars
- Percent values are expressed as decimal numbers in percent points (e.g., 12.5, not 0.125)
- Month values are integers
- Holder allocation:
cash_amount = upfront_cash × holder_pct, stock_amount = stock_value × holder_pct
Step 8: Produce the Final JSON
Follow the answer template exactly:
- Every field in the template must appear in the output, even if null
- Use stable IDs from the workbench (term IDs, consent IDs, contract IDs, employee IDs, finding IDs)
- Sort arrays as the template instructs (typically by issue_id, priority, or redline_id)
- Use only the allowed enum values
- Do not invent IDs or values not present in the template or workbench
- Do not include explanatory prose outside the JSON
Issue Classification Guide
Use these issue status values consistently:
| Status |
Meaning |
draft_exceeds_playbook |
Draft term gives counterparty more than the playbook's fallback allows |
draft_below_playbook |
Draft term gives the client less protection than the playbook requires |
missing_required_term |
A term the playbook or policy requires is absent from the draft |
out_of_policy |
Draft term exceeds a policy threshold requiring committee approval |
in_policy |
Draft term is within acceptable bounds — include only when template requires it |
Risk and Action Classification
Risk ratings:
HIGH: Direct dollar exposure above materiality thresholds, closing certainty at risk, or regulatory clearance required
MEDIUM: Moderate exposure, secondary protections missing, or procedural gaps
LOW: Minor gaps or notice-only items
Recommended actions:
revise: Modify an existing draft term toward the playbook position
add: Insert a missing required term
delete: Remove a draft term the playbook prohibits
accept: Term is within acceptable bounds
reject: Term is fundamentally unacceptable
approve / approve_with_conditions: Committee escalation context only
escalate: Needs higher-level business decision
Data Formatting Conventions
- Currency: integer dollars (no cents, no commas in JSON numbers)
- Percentages: decimal numbers in percent points to the precision specified by the template (typically two decimal places or one decimal place)
- Months: integers
- Dates:
YYYY-MM-DD strings
- Holder percentages: to four decimal places (e.g., 0.1850) when specified
- Empty arrays use
[], not null
- Missing source term IDs use
[] (empty array), not null
- Booleans are
true/false, never strings
Validation Checklist
Before delivering the final JSON, verify:
- The
deal_id matches the task
- Every required top-level field from the template is present
- All enum values come from the template's allowed set
- All dollar amounts are integers computed from the correct basis
- All stable IDs (term IDs, consent IDs, contract IDs, employee IDs) match workbench records exactly
- Array lengths match the template's expectations (no missing or extra items)
- The
priority_order array references only issue_ids that appear in the issue register
- Summary metrics are internally consistent (counts match array lengths, totals sum correctly)
- No narrative text appears outside the JSON structure
- All nulls are intentional — every null field in the output corresponds to a field that genuinely has no applicable data
Using the SQL Endpoint
When cross-table verification helps, use POST <BASE>/api/query with token deal-workbench-readonly. Keep queries to single SELECT or WITH statements. Use this sparingly — the dedicated API endpoints should cover most needs. SQL is most useful for verifying counts, cross-referencing IDs across endpoints, or aggregating values that span multiple records.
1---2name: deal-workbench-review3description: How to review M&A deal documents using the deal workbench API. Use when the task involves reviewing purchase agreements, producing issue registers, preparing closing packages, compiling committee escalation memos, conducting transition reviews, or building deviation matrices against a deal workbench. Applies to buyer-side or seller-side APA/SPA reviews, playbook comparisons, policy threshold checks, and any structured M&A deal analysis where the task provides an answer_template.json payload and references a running deal workbench at a TASK_ENV_BASE_URL.4metadata:5 source: https://skillmd.com/skills/prism-shadow/deal-workbench-review6---78# Deal Workbench Review910This skill teaches a systematic method for completing M&A deal review tasks against the deal workbench API. The core pattern is: read the template first, gather all deal data from the workbench, compare draft terms against playbook rules or policy thresholds, compute dollar amounts from the headline purchase price, and produce JSON that conforms exactly to the provided answer template.1112## Key Principle1314The task always provides an `answer_template.json` payload. This is your contract for output shape. Read it first and treat every field, enum, and constraint in it as non-negotiable. The workbench API provides the facts; the template defines what shape those facts must take.1516## Connecting to the Workbench1718The task always provides the base URL as `<TASK_ENV_BASE_URL>` (or similar placeholder). Substitute it everywhere. The read-only SQL endpoint, when available, is `POST <BASE>/api/query` with body `{"token": "deal-workbench-readonly", "sql": "<single SELECT or WITH statement>"}`.1920The primary API surface for a deal `DEAL_ID` consists of these endpoints:2122| Endpoint | Purpose |23|----------|---------|24| `GET <BASE>/workspace` | Workspace overview |25| `GET <BASE>/api/deals/<DEAL_ID>` | Deal record (headline price, parties, structure) |26| `GET <BASE>/api/deals/<DEAL_ID>/terms` | Current draft terms |27| `GET <BASE>/api/deals/<DEAL_ID>/risk-estimates` | Risk estimates for the deal |28| `GET <BASE>/api/deals/<DEAL_ID>/employees` | Employee data |29| `GET <BASE>/api/deals/<DEAL_ID>/consents` | Required third-party consents |30| `GET <BASE>/api/deals/<DEAL_ID>/regulatory` | Regulatory filings and status |31| `GET <BASE>/api/deals/<DEAL_ID>/benchmarks` | Market benchmarks for terms |32| `GET <BASE>/api/deals/<DEAL_ID>/documents` | Deal documents |33| `GET <BASE>/api/deals/<DEAL_ID>/notes` | Negotiation notes |34| `GET <BASE>/api/deals/<DEAL_ID>/cap-table` | Capitalization table (stock deals) |35| `GET <BASE>/api/deals/<DEAL_ID>/material-contracts` | Material contracts |36| `GET <BASE>/api/deals/<DEAL_ID>/diligence-findings` | Diligence findings |37| `GET <BASE>/api/playbooks` | Available playbooks |38| `GET <BASE>/api/playbooks/<PLAYBOOK_ID>/rules` | Playbook rules for a side |39| `GET <BASE>/api/policies` | Available policies |40| `GET <BASE>/api/policies/<POLICY_ID>/thresholds` | Policy thresholds |41| `GET <BASE>/api/search` | Search across the workbench |4243The older non-`/api` prefixed routes (`/deals/<id>`, `/playbooks`, `/policies`) return the same data and can be used interchangeably with the `/api` variants. Use whichever the task prompt references.4445## Systematic Workflow4647Follow these steps in order. Complete each step before moving to the next.4849### Step 1: Read the Answer Template5051Read `input/payloads/answer_template.json` in full. Map every field:5253- **Top-level required fields** and their types54- **Allowed enum values** for every constrained field (risk_rating, issue_status, recommended_action, etc.)55- **Stable identifier lists** (possible_issue_ids, stable_redline_ids, etc.) — these define the universe of expected outputs56- **Numeric constraints** (integer dollars, percent points to N decimal places, integer months)57- **Array element shapes** — the exact fields each object must have5859The template is your schema. Every null in the template marks a place you must fill from workbench data.6061### Step 2: Gather the Deal Record6263Fetch `GET <BASE>/api/deals/<DEAL_ID>`. Extract:6465- The deal ID (confirm it matches the task)66- The headline purchase price (this is your calculation basis for all dollar amounts unless a source explicitly states a different basis)67- The parties (target, buyer, seller names)68- The deal structure (stock purchase, asset purchase, merger)69- The client side implied by the task context (buyer-side or seller-side counsel)7071### Step 3: Gather the Terms7273Fetch `GET <BASE>/api/deals/<DEAL_ID>/terms`. These are the current draft terms the counterparty has proposed. Map each term to its stable ID. Pay attention to:7475- Indemnity caps, baskets, and survival periods76- Escrow amount, duration, and release triggers77- Financing conditions and reverse break fees78- Restrictive covenants (non-compete, non-solicit)79- Employee provisions (service credit, PTO, retention)80- Tax allocation and transfer tax splits81- Governing law and forum82- Consent conditions and material contract treatment83- Closing conditions and outside dates84- Materiality scrapes, knowledge qualifiers, and sandbagging8586### Step 4: Gather the Rules or Policies8788Depending on the task type:8990- **Playbook comparison tasks**: Fetch `GET <BASE>/api/playbooks/<PLAYBOOK_ID>/rules`. The playbook defines preferred and fallback positions for each term category. Compare each draft term against the playbook's preferred, fallback, and prohibited positions.91- **Policy threshold tasks**: Fetch `GET <BASE>/api/policies/<POLICY_ID>/thresholds`. The policy defines approval thresholds that draft terms must not exceed without committee escalation.9293### Step 5: Gather Supporting Records9495Fetch all endpoints the task references. Common supporting data:9697- **Employees** (`/api/deals/<ID>/employees`): headcount, PTO liabilities, service credit needs, WARN risks98- **Consents** (`/api/deals/<ID>/consents`): which third-party consents are required, which are notice-only99- **Regulatory** (`/api/deals/<ID>/regulatory`): HSR filing status, thresholds, effort standards100- **Benchmarks** (`/api/deals/<ID>/benchmarks`): market data for term comparison101- **Risk estimates** (`/api/deals/<ID>/risk-estimates`): quantified risk estimates102- **Material contracts** (`/api/deals/<ID>/material-contracts`): contracts requiring consent or notice103- **Cap table** (`/api/deals/<ID>/cap-table`): for stock deals, holder breakdowns104- **Diligence findings** (`/api/deals/<ID>/diligence-findings`): issues from due diligence105- **Documents** (`/api/deals/<ID>/documents`): relevant deal documents106- **Notes** (`/api/deals/<ID>/notes`): negotiation context107108Use the read-only SQL endpoint for cross-table checks when needed — send POST to `/api/query` with `{"token": "deal-workbench-readonly", "sql": "..."}`.109110### Step 6: Perform the Analysis111112This step depends on the task type. Common analyses:113114**Playbook comparison (buyer or seller APA/SPA review):**1151. For each draft term, look up the corresponding playbook rule1162. Classify the issue status:117 - `draft_exceeds_playbook`: draft is more favorable to counterparty than playbook's fallback (e.g., higher cap, longer survival)118 - `draft_below_playbook`: draft is less protective than playbook requires (e.g., lower escrow, missing fee)119 - `missing_required_term`: a term the playbook requires is absent from the draft120 - `in_policy` / `out_of_policy`: for policy threshold tasks1213. Determine risk rating (HIGH/MEDIUM/LOW) based on dollar exposure and closing impact1224. Determine recommended action (revise/add/delete/accept)1235. Calculate dollar deltas: `draft_amount - fallback_amount` or `fallback_amount - draft_amount`124125**Committee escalation (policy threshold):**1261. For each draft term, compare against the policy threshold1272. Exclude terms that are in-policy or stale1283. For out-of-policy terms, compute the delta between draft and threshold1294. Provide benchmark context and quantified exposure130131**Closing package / transition review:**1321. Identify missing provisions the client side requires1332. Classify each gap and provide the required position1343. Produce redlines showing what must change135136### Step 7: Compute Dollar Amounts137138All dollar amounts derive from the deal's headline purchase price unless a specific source (term, finding, risk estimate) explicitly states a different basis.139140- `amount = headline_price × (percent / 100)`, rounded to integer dollars141- Percent values are expressed as decimal numbers in percent points (e.g., 12.5, not 0.125)142- Month values are integers143- Holder allocation: `cash_amount = upfront_cash × holder_pct`, `stock_amount = stock_value × holder_pct`144145### Step 8: Produce the Final JSON146147Follow the answer template exactly:148149- Every field in the template must appear in the output, even if null150- Use stable IDs from the workbench (term IDs, consent IDs, contract IDs, employee IDs, finding IDs)151- Sort arrays as the template instructs (typically by issue_id, priority, or redline_id)152- Use only the allowed enum values153- Do not invent IDs or values not present in the template or workbench154- Do not include explanatory prose outside the JSON155156## Issue Classification Guide157158Use these issue status values consistently:159160| Status | Meaning |161|--------|---------|162| `draft_exceeds_playbook` | Draft term gives counterparty more than the playbook's fallback allows |163| `draft_below_playbook` | Draft term gives the client less protection than the playbook requires |164| `missing_required_term` | A term the playbook or policy requires is absent from the draft |165| `out_of_policy` | Draft term exceeds a policy threshold requiring committee approval |166| `in_policy` | Draft term is within acceptable bounds — include only when template requires it |167168## Risk and Action Classification169170**Risk ratings:**171- `HIGH`: Direct dollar exposure above materiality thresholds, closing certainty at risk, or regulatory clearance required172- `MEDIUM`: Moderate exposure, secondary protections missing, or procedural gaps173- `LOW`: Minor gaps or notice-only items174175**Recommended actions:**176- `revise`: Modify an existing draft term toward the playbook position177- `add`: Insert a missing required term178- `delete`: Remove a draft term the playbook prohibits179- `accept`: Term is within acceptable bounds180- `reject`: Term is fundamentally unacceptable181- `approve` / `approve_with_conditions`: Committee escalation context only182- `escalate`: Needs higher-level business decision183184## Data Formatting Conventions185186- Currency: integer dollars (no cents, no commas in JSON numbers)187- Percentages: decimal numbers in percent points to the precision specified by the template (typically two decimal places or one decimal place)188- Months: integers189- Dates: `YYYY-MM-DD` strings190- Holder percentages: to four decimal places (e.g., 0.1850) when specified191- Empty arrays use `[]`, not `null`192- Missing source term IDs use `[]` (empty array), not `null`193- Booleans are `true`/`false`, never strings194195## Validation Checklist196197Before delivering the final JSON, verify:1981991. The `deal_id` matches the task2002. Every required top-level field from the template is present2013. All enum values come from the template's allowed set2024. All dollar amounts are integers computed from the correct basis2035. All stable IDs (term IDs, consent IDs, contract IDs, employee IDs) match workbench records exactly2046. Array lengths match the template's expectations (no missing or extra items)2057. The `priority_order` array references only issue_ids that appear in the issue register2068. Summary metrics are internally consistent (counts match array lengths, totals sum correctly)2079. No narrative text appears outside the JSON structure20810. All nulls are intentional — every null field in the output corresponds to a field that genuinely has no applicable data209210## Using the SQL Endpoint211212When cross-table verification helps, use `POST <BASE>/api/query` with token `deal-workbench-readonly`. Keep queries to single SELECT or WITH statements. Use this sparingly — the dedicated API endpoints should cover most needs. SQL is most useful for verifying counts, cross-referencing IDs across endpoints, or aggregating values that span multiple records.