#reproducibility

How fast can a coding agent p-hack? Measured: on known-zero data, realistic searches manufacture p<.05 on 33-97% of null draws, in ~1 second each. Instrumented specification-search engine + 11 Claude Code skills for econometric designs; every search leaves a verifiable ledger and an honest p-value. Research and teaching use only.

Release repository for agent benchmark evidence-reporting artifacts and reproduction workflows.

Controlled coding-agent harness ablations with auditable traces, reproducible evidence bundles, and SWE-bench interoperability.

Standard to define citable scientific protocols for use by AI agents

一个 7×24 小时不间断运行，帮你跑实验、用 Git 维护公开仓库、更新 LaTeX（Overleaf）论文的科研自动化工作流。实测效率提升 25 倍。Claude Code Skill，一条命令装进任意项目。

Evidence-backed scientific AI report infrastructure

Executable biomedical workflows for AI assistants — literature analysis, biomarker discovery, and Mendelian randomization through reproducible agent skills.

Verifiable run bundles for AI and agent executions: tamper-evident evidence, first-class redaction, semantic diffs for CI regression

Tests exact agent interventions—skills, prompts, models, tools, and workflows—and publishes what changed, what failed, and what decision follows.

Research enhancement methodology skill — pre-registration, provenance, determinism, adjudication, formal verification, audit loops, knowledge distillation. Agent-executable research-integrity workflow (SKILL.md convention); methodology only, zero theory detail.

A local-first desktop interface for chatting with locally managed AI models and, progressively, for querying and exploring repositories of scientific dynamical models.

Rust tool to compile project facts and external sources into Codex, Claude Code, GitHub Copilot, and generic Agent configurations.

Estimand-indexed audit of LLM synthetic respondents against CGSS benchmarks

Trustworthy, reproducible AI data analysis in Claude Code. A data-science plugin (skill + slash commands) that stops your coding agent building on stale data and records what produced every result.

Measuring and verifying LLM decision agents under orbital dynamics, delay, and intermittency: a reproducible testbed, failure-mode taxonomy, and in-chain verification.

A provable ceiling on LLM routing, voting and mixture-of-agents, measured on 67 models: code, audit and one-command reproduction

Independent Comparator/Lean audit of two OpenAI mathematical results, with bilingual evidence, scope notes and an essay on verification and explanation.

OpenSIEM-AI: reproducible open-source SIEM with ML detection and automated SOAR containment (research code)

Skills, hooks, and CLI checks for safer biomedical AI coding agents.

Transparent benchmark cards, JSON artifacts, and visual tools for photonic AI accelerator energy/noise claims.

Run provenance, comparison validity, and the shared primitives for agentic work on research infrastructure. A base layer, not an agent framework.

“between models” agent handoff protocol. late 2025. deprecated.

Agent-first robotics skill discovery: pinned source examples, reproducible preparation, and reviewed task findings.

Measures the gate of a self-improving image-restoration loop: 50% false positives, and a ten-line baseline beats the LLM agent. CPU-only.

Decode papers. Adapt experiments. Learn by running them anywhere.