xhowenma/autodojo
Code for "AutoDojo: Adaptive Black-Box Attacks Reveal the Limits of IPI Defenses and Task-Specification Effects in LLM Agents"
GITHUB SIGNALS / PROJECT INDEX
Search the archive. Follow your curiosity. Keep the projects you want to come back to.
Code for "AutoDojo: Adaptive Black-Box Attacks Reveal the Limits of IPI Defenses and Task-Specification Effects in LLM Agents"
A cross-platform Typst editor.
Open the project to explore the idea and the code.
An open-source research agent system for your Zotero library.
Rust-based waybar widget to monitor status of Claude, GPT, GLM, OpenRouter plans/credits - inspired by claudebar/codexbar
Pure-Rust, dependency-lean, ultra-fast Markdown -> beautiful all-in-one HTML & tiny optimized PDF (library + single-binary CLI: fmd)
A lightweight library for xPU kernel JIT compilation
A native terminal that orchestrates coding agents. Claude Code and Codex keep their own TUI; Runner adds sessions, skills, and crews.
Your personal AI agent platform — self-hosted, BYO Claude/Codex subscription
Open the project to explore the idea and the code.
Official implementation of Gradient-Update Alignment (GUA) for conflict-free training of Physics-Informed Neural Networks (PINNs).
The PyTorch for code world models: build and deploy world models in <2 minutes (100+ worlds, 3 papers).
An open collection of 160+ Legal AI Skills for ChatGPT, Claude, Grok, Cursor & Gemini.
Consort keeps AI-written code clean and correct: a coordinated ensemble of role agents, spec-first and test-driven, driven by a deterministic state machine against live branches of a real Lakebase database. Engineering discipline is no longer left to the whim of a model. Consort enforces it.
Offline context handoff between coding agents. Read local histories into fresh sessions; source stores stay unchanged. Stdlib-only Python, not live session restore.
AI agent skills for social media content strategy, creation, and analysis across text-first platforms
Penpot AI kit.
Open measurement infrastructure for agent-facing software. Separate operations from attempts, evidence from inference, and usage from value. Reach → Choice → Use → Utility → Value.
Open the project to explore the idea and the code.
Subscription-powered Fable and Astra research for optimization solvers, with isolated experiments, independent verification and human-reviewed evidence.
AI Systems as Scientific Instruments for Understanding the Mechanisms of Intelligence
Twenty LuaLaTeX and MetaPost (fiziko) examples of the old scientific-paper aesthetic: engraved plates, star charts, computed log tables, knots, telescopes
Programmable World Model: explicit world state, generative rendering.
How to train your slop cannon: a short guide to using large language models for research in the natural sciences (Loader, Oppenheim, Osborne)