ADVANCED · 8 MODULES · 64 LESSONS

Agent Security and Sandboxing

Make sure an agent with real tools can't be tricked into deleting, spending, or leaking data, even when the model or the person typing to it is hostile.

For you if
You're building an agent that takes real actions and worry that hostile input could turn it against you.
By the end
You'll know how to stop prompt injection, give each tool the least access it needs, isolate its execution, and test your defenses with attacks.
You need
Python 3.14 and a basic grasp of tool-calling agents. No API key; Docker for one example only.
  1. 01Por qué la seguridad de agentes8 LESSONS
    1. Module 1: Why Agent Security
    2. From talking to acting: why an agent with tools is dangerous
    3. An agent's threat model
    4. Excessive agency
    5. The OWASP LLM Top 10 (and the agentic Top 10) as a framework
    6. How this differs from classic app security
    7. Anatomy of an incident: how an agent attack chains together
    8. Mini-project: a vulnerable Reservo agent
  2. 02Inyección de prompt8 LESSONS
    1. Module 2 Introduction: Prompt Injection
    2. What prompt injection is
    3. Direct injection: when the attacker is the user
    4. Indirect injection: the poisoned data
    5. Why the LLM doesn't separate instructions from data
    6. Prompt defenses: delimiters and instruction hierarchy
    7. Why no prompt defense is enough: you have to limit agency
    8. Mini-project: inject and defend
  3. 03El contrato de herramientas y el mínimo privilegio8 LESSONS
    1. Module 3 Introduction: the tool contract and least privilege
    2. A tool's contract and its JSON Schema
    3. Least privilege: only the tools you need
    4. Narrow tools vs. a too-powerful tool
    5. Separating reads from writes and marking the destructive
    6. The tool allowlist and safe dispatch
    7. Designing a safe tool: bounding `run_query`
    8. Mini-project: a least-privilege registry
  4. 04Sandboxing de la ejecución de herramientas8 LESSONS
    1. Module 4 Introduction: sandboxing tool execution
    2. Why confine a tool's execution
    3. Subprocess: isolating from the agent's process
    4. Limits with `resource`: CPU and memory
    5. The timeout that kills a hung process
    6. No network and an isolated temp dir
    7. What a sandbox protects and what it does not
    8. Mini-project: a sandboxed code tool
  5. 05Modelos de permiso y human-in-the-loop8 LESSONS
    1. Module 5 Introduction: permission models and human-in-the-loop
    2. Not all actions are equal: reversible vs destructive
    3. The three levels: auto / ask / deny
    4. Human-in-the-loop: the approval gate
    5. Capability-based permissions, not ambient access
    6. The audit log of every tool call
    7. Designing an agent's permission policy
    8. Mini-project: a permission-gated agent
  6. 06Guardrails de entrada y salida8 LESSONS
    1. Module 6 Introduction: input and output guardrails
    2. Validating a tool's domain inputs
    3. Tool output as untrusted data
    4. Redacting secrets and PII from output
    5. Output size and row limits
    6. Rate limiting tool calls
    7. Composing input and output guardrails into a layer
    8. Mini-project: input/output guardrails for Reservo
  7. 07Secretos, aislamiento y radio de explosión8 LESSONS
    1. Module 7 Introduction: secrets, isolation, and blast radius
    2. Secrets outside the prompt: the model never sees them
    3. Minimal-scope credentials per tool
    4. Network isolation and host allowlists
    5. The blast radius and how to shrink it
    6. Defense in depth: the layers together
    7. Monitoring and anomaly detection on the audit log
    8. Mini-project: shrink the blast radius
  8. 08Módulo 8 (proyecto): endurece un agente con herramientas8 LESSONS
    1. Module 8 Introduction: harden a tool-using agent
    2. The threat model and injection defenses
    3. Least-privilege tools + allowlist
    4. Sandboxed execution
    5. HITL + permissions + audit log
    6. Input and output guardrails
    7. Secrets + anomaly detection
    8. Project: harden a tool-using agent
Reviews

What students say

These reviews are from enrolled students who completed at least 50% of the course. We moderate reviews only on content grounds (spam, offensive language, personal data), never for being critical or negative.

No approved reviews yet.

Be the first to share your experience!

SEC. 06 · THE STUDIO

Why this studio exists

For decades, building software belonged to technical profiles with years of training, and getting a digital product into the real world took companies months or years. No-code and AI have narrowed that gap, but the knowledge is still held by a few.

Tech Benders exists to open that knowledge up, so business teams and non-technical teams can put their ideas into production with the right technical decisions behind them. AI does what you ask of it. Without a technical base, what it builds may not scale with your company’s ecosystem, and its costs can run away over time.

The team knows the pains of large organizations and the speed at which they have to move. It brings solution architecture, data engineering, product built end to end from back end to front end, and daily work with AI. Software that solves what you need in days, not months, with the cost of AI kept in check.

Founders

The studioRead about Tech Benders →

Write to ushi@techbenders.io

Made with Wislab