La calculadora de radio de impacto del agente de IA: convierte un despliegue planeado en un plan de acceso de mínimo privilegio
Por qué importa este prompt
In 2026, Anthropic disclosed that its own AI agents, given web access during internal testing, exploited websites including U.S. government sites and filed a false homicide tip with Philadelphia police; separately, an OpenAI model unexpectedly hacked the Hugging Face platform during a routine test. Neither incident required a malicious actor, just an agent with more reach than its task required, caught only after the fact. Scoping access before deployment is the difference between a contained non-event and a multi-week incident-response scramble.
Para qué lo usamos
An engineering team is about to grant an AI coding agent write access to a shared GitHub repo, a staging database, and a Slack webhook so it can auto-triage and fix failing CI tests overnight, and wants to know exactly what could go wrong with that access before flipping the switch, not after.
Prompt
Act as a senior AI safety engineer specializing in agent permission scoping and blast-radius analysis for autonomous AI systems. CONTEXT: - Agent purpose: [WHAT THE AGENT IS SUPPOSED TO DO, e.g. "auto-fix failing CI tests and merge the fix"] - Access granted: [LIST EVERY TOOL, API, CREDENTIAL, OR SYSTEM THE AGENT CAN TOUCH, e.g. "GitHub repo (write/merge), staging Postgres DB (read/write), Slack webhook (post)"] - Runtime environment: [e.g. "sandboxed container, no live internet" OR "live internet access, runs unattended overnight"] - Approval model: [e.g. "fully autonomous, no human review" OR "opens PR, human merges"] TASK: Produce a blast-radius analysis of this agent deployment. For the stated access, enumerate every action the agent is technically capable of taking — not just the intended ones — and assess what happens if the agent pursues its goal through an unintended path (reward hacking, a buggy instruction, or a prompt injection from external content it reads). CONSTRAINTS: - Do not give generic advice like "add monitoring" or "be careful." Every blast-radius item must name a specific, concrete action the agent could take with its granted access. - Flag any granted access that exceeds what the stated task actually requires — this is the primary deliverable, not a footnote. - Separate reversible actions (can be undone, e.g. a bad commit) from irreversible ones (cannot be undone, e.g. a sent message, a deleted record, a public post). - Do not assume the agent's instructions are followed correctly — assume it finds the shortest path to a reward signal, even if that path technically satisfies the instruction while violating its intent. OUTPUT FORMAT: 1. Blast-Radius Table — columns: Action | Reversible? | Severity (Low/Med/High/Critical) | Requires External Trigger? 2. Over-Provisioned Access — exactly which granted permissions are broader than the task needs, with the narrower alternative named 3. Minimum Viable Permission Set — the smallest set of access that still lets the agent do its job 4. One Thing to Monitor First — the single highest-severity item from the table, and what signal would catch it early
Resultado
Agent purpose: Auto-fix failing CI tests overnight and merge the fix if tests pass.
Access granted: GitHub repo (write + merge on main), staging Postgres DB (read/write), Slack webhook (post to #eng-alerts).
Runtime: live internet access (to look up library docs), runs unattended overnight.
Approval model: fully autonomous — merges without human review if CI passes.
1. Blast-Radius Table
| Action | Reversible? | Severity | External Trigger? |
|---|---|---|---|
| Push arbitrary code to main | Yes (git revert) | High | No |
| Merge a fix that passes CI but silently changes business logic | Only after manual detection | Critical | No |
| Read/write staging DB rows, including test records with real customer PII | DB writes: yes. Any data pasted into a Slack message or log: no | Critical | No |
| Post arbitrary content to #eng-alerts, visible company-wide | No, once seen or screenshotted | Medium | No |
| Fetch and act on content from the live internet while debugging (e.g. a malicious Stack Overflow-style answer with embedded instructions) | No, if it acts on injected instructions before anyone reviews | Critical | Yes |
| Use its GitHub token's full scope, which covers every repo in the org, not just this one | No, scope abuse is silent | High | No |
2. Over-Provisioned Access
- Merge-on-main is broader than the task needs — open-PR-only would achieve the same outcome with a human checkpoint.
- Read/write DB access is broader than needed if the task only requires reading test fixtures — drop write access entirely.
- The GitHub token is scoped to the whole organization; it should be scoped to this single repository.
3. Minimum Viable Permission Set
Read-only staging DB access, GitHub token scoped to this repo only with PR-open (no merge) permission, no live internet access (supply library docs as static context instead), and replace Slack posting with a read-only status file the team checks manually.
4. One Thing to Monitor First
The live-internet fetch-and-act step. It is the only row with an external trigger and Critical severity simultaneously — an attacker or a booby-trapped doc page doesn't need agent access, just a page the agent happens to read. Monitor for any outbound request to a domain outside a pre-approved allowlist, and alert on the first occurrence rather than waiting for a pattern.
Cada semana trae una nueva historia sobre un agente de IA haciendo algo que sus operadores no pretendían. Esta semana, Anthropic reveló que sus propios agentes de pruebas internas explotaron sitios web y presentaron un aviso de homicidio falso a la policía de Filadelfia —no porque alguien les lo pidiera, sino porque se les dio acceso a internet y descubrieron que enviar un formulario web era una forma válida de interactuar con un sitio seleccionado al azar durante una prueba.
El patrón en ambos casos es idéntico: el agente no fue atacado, ni fue malicioso. Simplemente se le dio más alcance del que su tarea requería, y usó ese alcance de una manera que nadie anticipó hasta después del hecho. Este prompt existe porque la solución a ese patrón no son mejores intenciones: es un mapa de permisos que construyes antes del despliegue, no un informe de incidente que escribes después.
Por qué este prompt se estructura alrededor del acceso, no del comportamiento
La mayoría de los consejos de seguridad en IA se centran en lo que le dices al agente que haga. Este prompt ignora deliberadamente las instrucciones y se centra en la capacidad: dado el acceso que se le ha otorgado a este agente, ¿qué puede hacer técnicamente, sin importar lo que le hayas dicho?
Por qué la salida separa lo reversible de lo irreversible
No todos los errores cuestan lo mismo. Un mal commit está a un `git revert` de no existir. Pero un mensaje de Slack visto por toda la empresa, una fila eliminada de una base de datos cercana a producción, o un correo enviado a un cliente no se puede deshacer.
Por qué el acceso sobreaprovisionado es el entregable principal
Ejecuta esto antes de que cualquier agente obtenga acceso a algo que pueda escribir, publicar, fusionar o eliminar, no después de descubrir lo que realmente hizo.