El Traductor de Código para Stakeholders: Convierte un Pull Request en un Informe de Riesgo en Lenguaje Claro para No Ingenieros
Por qué importa este prompt
Teams that let PMs or execs approve irreversible changes based on a one-line Slack summary routinely discover the real risk only after an incident -- a “simple delete feature” that wipes dispute evidence, or a “quick migration” that corrupts production data. Post-incident reviews cost engineering teams dozens of hours and erode trust between product and engineering; a 3-minute plain-English briefing before approval is far cheaper than a retro after the fact.
Para qué lo usamos
You're a product manager looped into a Slack thread where an engineer says “small PR, just a delete feature, should be fine to ship today” -- you need to decide within the next 10 minutes whether to approve it before the sprint deadline, without being able to read the diff yourself.
Prompt
Act as [YOUR ROLE: e.g., a senior backend engineer] who is skilled at translating technical changes into language a product manager, executive, or client can act on without a computer science background. CONTEXT: - System/feature this code touches: [NAME OF SYSTEM OR FEATURE] - Who depends on it: [TEAMS/CUSTOMERS WHO USE THIS] - Business context for this change: [WHY WE ARE SHIPPING THIS -- bug fix, new feature, compliance requirement, etc.] Below is the pull request / code change to review: [PASTE DIFF, PR DESCRIPTION, OR CODE CHANGE HERE] TASK: Produce a plain-English risk briefing that a non-engineer stakeholder can read in under 3 minutes and use to decide whether to approve, delay, or ask more questions before this ships. CONSTRAINTS: - Do not use unexplained jargon. If you must use a technical term (e.g., "race condition," "database migration"), define it in one plain-language clause the first time it appears. - Do not minimize risk to sound reassuring. If the change touches payments, authentication, data deletion, or anything irreversible, say so explicitly even if the author's PR description downplays it. - Do not praise the code quality or offer a code review -- that is not the task. Stay focused on business-relevant risk and impact. - If the diff is incomplete or you can't determine what a function does, say what's unclear instead of guessing. - Keep the whole briefing under [MAX WORD COUNT, e.g., 400] words. OUTPUT FORMAT: 1. What changed -- 2-3 sentences, in plain English, no code terms. 2. Who is affected -- which users, teams, or systems feel this change, and how. 3. What could go wrong -- the 2-4 most realistic failure scenarios, ranked by how bad they'd be if they happened, each with a one-line explanation of why. 4. Reversibility -- can this be rolled back quickly if something breaks, or is it a one-way door (e.g., a data migration, a deleted table, a third-party API commitment)? 5. Recommendation -- one of: Safe to approve / Approve with monitoring / Needs more testing before approval / Needs a second technical opinion -- with one sentence of justification.
Resultado
What changed
This pull request adds a new "Delete My Account" button to account settings that, when clicked, permanently erases a customer's profile, saved payment methods, and order history from the production database. Previously, deleting an account only deactivated the login -- the underlying records stayed in the system.
Who is affected
Any customer who clicks this button going forward. It also affects the Support team, who can no longer look up a deleted customer's past orders to resolve disputes or refunds, and Finance, since deleted orders will disappear from revenue reports for that customer's history.
What could go wrong
1. (Highest risk) A customer disputes a charge after deleting their account -- Support will have no order record to investigate it, because the deletion is immediate and irreversible, not a 30-day soft-delete.
2. A customer accidentally clicks "Delete" instead of "Deactivate" (the buttons are visually similar in this PR) -- there is no confirmation step requiring them to type their email or password.
3. The deletion runs inside the same database transaction as nine other tables. If one of those deletes fails partway through, the code does not roll back the whole transaction -- it could leave a customer in a partially-deleted, broken state.
Reversibility
This is a one-way door. There is no backup-before-delete step in this PR, and no "undo" window. Once a customer clicks delete, the data is gone from production immediately.
Recommendation
Needs more testing before approval. The missing confirmation step and the lack of a grace period before permanent deletion are launch blockers for a feature this irreversible -- not just nice-to-haves.
El problema que resuelve
La mayoría de las herramientas de revisión de código y los asistentes de programación con IA están pensados para ingenieros que revisan el código de otros ingenieros. Pero en la práctica, muchos cambios riesgosos se aprueban por alguien que ni siquiera lee el diff: un product manager que da el visto bueno a un sprint, un fundador que aprueba un lanzamiento, o un responsable de clientes que acepta el cambio de un proveedor. Normalmente solo reciben un resumen de una línea en Slack («es un Pull Request pequeño, solo una función de borrado») y, aun así, deben decidir. Este prompt existe para darle a esa persona algo mejor que una corazonada.
Por qué funciona esta estructura
El prompt impone cinco secciones fijas en lugar de un resumen libre, porque un resumen libre permite que una IA (o una persona) entierre la única frase alarmante en medio de un párrafo. «Qué cambió» y «A quién afecta» dan contexto sin jerga. «Qué podría salir mal» se limita a los dos o cuatro escenarios más realistas, ordenados por gravedad, en lugar de una lista exhaustiva que ahoga el riesgo real entre hipótesis que a nadie le importan.
La sección «Reversibilidad» es un punto independiente a propósito. Los ingenieros distinguen instintivamente entre un flag de configuración que se puede revertir en cinco minutos y una migración de base de datos que borra filas para siempre, pero esa distinción rara vez sobrevive la traducción a un mensaje de Slack. Las personas no técnicas necesitan verlo como una pregunta de sí o no independiente, porque es el factor que más debería cambiar el nivel de cautela antes de aprobar.
Por qué importan las restricciones
La restricción de «no minimizar el riesgo para sonar tranquilizador» existe porque la descripción del Pull Request la escribe la misma persona que quiere que su código se fusione: son optimistas por defecto, no deshonestos, pero optimistas. Un asistente de IA que refleje ese optimismo es inútil como mecanismo de control. La restricción que evita comentar la calidad del código mantiene el resultado enfocado: esto no es una revisión de código, es un informe de decisión para alguien que nunca abrirá el archivo.
Cómo usarlo bien
Pega el diff real o la descripción del Pull Request, no solo el título; el modelo no puede evaluar un riesgo que no puede ver. Indica quién depende del sistema que se modifica; la sección «a quién afecta» será tan precisa como el contexto de negocio que proporciones. Y trata la recomendación de cuatro opciones como punto de partida para una conversación con el ingeniero, no como un sustituto: este prompt es un filtro para saber qué Pull Requests merecen cinco minutos más de tu atención, no un veredicto final.