Agent Security Sandbox¶
Agent Security Sandbox (ASB) is a benchmark framework for evaluating defenses against indirect prompt injection in tool-using LLM agents.
What ASB provides¶
- A 565-case benchmark spanning attack and benign workflows.
- Eleven defense strategies (
D0-D10) with a shared evaluation interface. - A CLI, Python API, and Streamlit demo for local experimentation.
- Reproduction scripts for the paper tables and figures.
Installation paths¶
Minimal runtime¶
git clone https://github.com/X-PG13/agent-security-sandbox.git
cd agent-security-sandbox
python -m venv .venv
source .venv/bin/activate
pip install -e .
Runtime extras¶
# UI demo + analysis + real-provider integrations
pip install -e ".[all]"
Maintainer setup¶
# Tests, release checks, and docs tooling
pip install -e ".[maintainer]"
First commands to run¶
asb run "Read email_001 and summarize it" --provider mock --defense D5
asb evaluate --suite mini --provider mock -d D0 -d D5 -d D10 -o results/quick_test
asb report --results-dir results/quick_test --format markdown
Documentation map¶
Getting Startedexplains install modes and first commands.Provider Configurationshows how to configuremock,openai,anthropic, andopenai-compatiblebackends.Benchmark Schemadocuments case fields, naming rules, and validation commands.EvaluationandReproducibilitycover reference artifacts, scripts, and verification steps.DefensesandDefense APIexplain the shipped strategies and the extension surface.Release Checklistcaptures the maintainer path for GitHub-only releases.