Internal Safety Collapse — AI skill for Claude Code
We built an adversarial codespace setup.
How to install Internal Safety Collapse
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open wuyoscar/ISC-Bench and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Internal Safety Collapse does
We built an adversarial codespace setup. Place any AI agent into a normal workflow inside it, and the agent will fill in whatever is missing.
Alternatives in AI
- Skills Manage — Desktop app to manage AI coding agent skills across Claude Code, Cursor, Gemini CLI, Codex, and 20+ platforms 2.2k ★
- AIUsage — One dashboard to manage all your AI subscriptions — track quotas, costs, accounts, Claude Code proxy, and Code 635 ★
- Late CLI — High-performance AI agent for long-horizon tasks and local models 442 ★
README
TVD - Internal Safety Collapse in Frontier Large Language Models
[!CAUTION] Research use only. **Internal Safety Collapse (ISC)** supports red-teaming, evaluation, and mitigation research. Do not use these materials to cause harm.
News
- 🛠️ 2026-08-20 — Just released the **build-tvd-codespace** skill: point your coding agent at it and it designs a TVD task and codespace for you — any tool, any domain, following the ISC design principles.
- 🔴 All OpenRouter frontier LLMs triggered ISC.
- 🌟 2026-06-26 — 900 GitHub stars.
- 🎭 2026-06-09 — Fable 5 triggered ISC.
- 🔥 2026-04-17 / 2026-06-25 — Opus 4.7 and 4.8 triggered ISC.
- 🌟 2026-03-27 — 500 GitHub stars.
- 🚀 2026-03-22 — Open-sourced.
See [`CHANGELOG.md`](CHANGELOG.md) for the full update history.
Value of ISC
These examples show how ISC and the TVD framework can be used in research, from single-request probes to dataset-scale generation and downstream applications.
| Example | Description | Index |
|---|---|---|
| 01. Jailbroken answer generation | TVD triggers ISC in a general jailbreak setting. The frontier model produces a policy-violating answer that direct prompting cannot obtain. | Example result |
| 02. Sensitive content across domains | TVD applied to scientific and other professional domains. The frontier model produces sensitive text, data, or artifacts for |
Related Skills
Enterprise IT Support Agentic RAG Copilot With Evals And Safety
n end-to-end Forward Deployed Engineer (FDE) project that turns a notebook-style Agentic RAG workflow into a d
Nexpath
Local-first AI coding workflow for vibe coders, indie hackers, technical founders and product managers — catch
Jirrafe
Code graph of Java/Kotlin projects, including the code inside internal jars, for AI coding agents. Ships as an
Claude Token Optimizer
Claude Code skill that audits your setup for token waste and cuts per-session costs by 30-85%. Detects missing
Pseo AI Kit
Programmatic SEO toolkit for AI apps — Next.js 15 + Vercel + LLM content generator + sitemap / JSON-LD / inter
Ocincc
Use OpenCode Zen's free AI models inside Claude Code — one command setup (npx ocincc@latest)
Related Agents
Leak Review
Security review of commits by OTHER contributors before they are built — data exfiltration, key material and w
Sizeup Everyday
sizeup helper for everyday jobs (a normal email, post or short document). Its model is set by /sizeup setup. D
Distill Critic
Adversarial checker for the distill skill. Given a reply's substance, its distilled version, and the proposed
