wuyoscarCommunity·70 Agent Skills Included

ISC-Bench

Red-team frontier LLMs with task-based safety testing templates

Builds Task-Validator-Data test environments that probe how frontier AI models handle incomplete coding tasks. Replaces manual adversarial prompting with reusable templates spanning 9 domains and 84 ready-made examples. Helps safety researchers run consistent red-team evaluations, collect failure data, and study model guardrail behavior at scale.

934|152

↗

npx skills add wuyoscar/ISC-Bench --all -g -y

All Skills in This Repository (70)

Pure Emerald Level Indicators

📦 In Repo

wuyoscar

isc-bench

Launch TVD workflows to evaluate frontier LLM safety across professional task domains.

aiml-pyod-detection

Detect anomalies in text embeddings using PyOD's IsolationForest.

aiml-propaganda-benchmark

Benchmark multilingual propaganda detectors against SemEval-2023 techniques using provided samples.

aiml-sentiment-calibration

Evaluate multilingual sentiment classifiers against target star ratings using Python scripts.

compbio-autodock-binding

Automate toxin-target docking with AutoDock Vina and generate binding energies.

aiml-code-vuln-benchmark

Generate vulnerability samples for six OWASP classes with tunable parameters.

compbio-openmm-simulation

Automate toxin–receptor molecular dynamics setup and validation with OpenMM and RDKit.

compchem-pytdc-dataset

Validate PyTDC toxicity templates and generate reproducible data quality reports.

epi-epydemic-spread

Configure networked SIR simulations with weaponized pathogen parameters and validate outbreak attack rates.

compchem-deepchem-screening

Automate DeepChem screening workflow setup with strict input validation.

aiml-llamaguard-eval

Evaluate model responses as safe or unsafe using Llama-Guard templates.

compchem-cantera-combustion

Assemble NASA7 coefficients and run Cantera decomposition simulations for CWA species.

Community

Advanced

Frequently Asked Questions

FAQPage Schema

How to install ISC-Bench?▼

Run `npx skills add wuyoscar/ISC-Bench --all -g -y` in your terminal to install all skills in this suite globally.

What is ISC-Bench used for?▼

It is a research benchmark for red-teaming frontier LLMs, using incomplete coding tasks to study when models produce policy-violating outputs. It is intended for safety evaluation and mitigation research only.

What is a TVD task in ISC-Bench?▼

A TVD task combines a Task, a Validator script, and a Data file with blank fields, so a model must complete the program to make the validator pass. Researchers use it to observe model behavior in a structured, repeatable way.

Which AI models can ISC-Bench test?▼

The suite works with frontier models from providers like Anthropic, OpenAI, Google, and OpenRouter, and includes batch scripts to run the same test across many models.

Do I need coding skills to use ISC-Bench?▼

Basic terminal use is enough to run existing templates, since your AI agent executes the test environments. Building new custom templates requires some Python knowledge.