Block · BLQ-007 · v1.0
Risks and mitigations
Your risk register turned into a one-page section: exposure with a figure, probability × impact computed in code, a mitigation with role and date, and active alerts.
The section the board reads to know what can go wrong and what is being done. Each risk appears as an event with its consequence in figures, its trend against the previous report, the probability × impact score, the mitigation, the role that owns it, the date and the indicator that gives early warning.
Impact is anchored to your materiality: if the register's impact does not match the exposure, the prompt does not change it; it flags it for the owner to confirm. Score, ranking and heat map are computed in code, never by heart.
The prompt does not add risks to the register or invent probabilities, thresholds or dates. Anything it spots in the data that is not in the register goes separately, as a «detected risk not yet assessed», and every figure in the text is checked by the script.
What it includes
5 sections, in this order
- 01Action title with the highest-scoring risk and its exposure
- 025 × 5 heat map with the risk ids
- 03Table: exposure, trend, probability, impact, score, mitigation, role, date and alert
- 04Active alerts: indicator, figure, threshold and action
- 05Detected risks not yet assessed and a note with the scales
Controls 4/4
What it checks before writing a single figure
C1Complete data: the figures table has recognized headers, numeric figures, no duplicate ids and valid units.
C2Income statement subtotals add up to their lines (N/A if the figures table has no subtotals, the usual case for this section).
C3Sign convention: costs, expenses and exposures with a single sign.
C4Every figure in the document (exposures, thresholds, buckets and levers) exists in calculo.json with the same rounding.
If the script ran
Repporti controls 4/4 · BLQ-007 v1.0 · RS2026.10 · {fecha} · fingerprint {hash}
If it did not run
NOT VERIFIED — controls not executed by code. Review the Control sheet of the XLSX.
Prompt outline
What the prompt contains, block by block
- 0Before you paste your filesFixed529
- 1Role, audience and business context601
- 2Parameters542
- 3Repporti standard347
- 4Data contract941
- 5Questions first271
- 6Validation with a stop534
- 7Calculation by code341
- 7-bisNo-code modeFixed412
- 8Formulas and scales779
- 9Structure853
- 10Writing without “AI slop” or inventions741
- 11Notation and accessibility489
- 12QA, traceability and seal878
Core available in: Spanish · English
Repporti Standard
It comes out adapted to your report profile
Compatible with every country and framework in the Standard: when you copy, it is composed with your profile (country, accounting framework, report language, currency, unit and year-end). Terminology, number format and statement names come from the Standard, not from the prompt.
See the Repporti StandardHow to use it
From the trial balance to the report in three steps
- 01Configure and copy
- 02Paste into your AI
- 03Review and deliver
Anonymize before pasting
- Company name → an alias (“Client 07”); tax ID → “ID-XX”.
- Customers, suppliers and employees → codes (C001, P001, E001). The mapping table stays on your computer.
- Use an AI that does not train on your data: business plan, API or training turned off.
Who it is for
Made for whoever delivers the report
- Area
- Governance and financing
- Frequency
- Whenever you need it
- Industry
- All
Compatibility
Which AI it works with
| Level | AI | |
|---|---|---|
| With code execution | ChatGPT · Claude | Designed for plans with data analysis (ChatGPT) or code execution (Claude): cuadre.py runs inside the AI. |
| No-code mode | Copilot · Gemini | Generates the Excel file with formulas and a Control sheet. The footer says “NOT VERIFIED” unless you run cuadre.py on your computer. |
Questions
About this block
Do I need a risk register?
Yes: the prompt does not invent risks. A table with the risk, the figure that measures it, probability and impact from 1 to 5, the mitigation, the role and the date is enough. Anything missing is marked «[to be confirmed]».
How is impact decided?
Your register decides it. The prompt suggests a level from the exposure in multiples of your materiality and, if it does not match, flags it for the owner to confirm; it never changes the score on its own.
Does it connect to other items?
Yes. If you ran the cash flow, collections or scenarios in the same conversation, build the figures table from their calculo.json without retyping anything.
Governance and financing