VCVinh

6 mods across 1 repository, 1 stars between them.

pipelineengine

01

VCVinh/Awesome-Research-Development

Command Claude Code

Load and apply the research-pipeline-verification-gates skill (17 protocols — verification, ontology, scale, comprehension depth, idea exhaustiveness) to a specified pipeline, project, or research task.

1 1mo ago A 0 tokens

VCVinh/Awesome-Research-Development

Skill Claude CodeCodex

Where the delegation boundary sits between what Claude can decide/do and what only Vinh can know or authorize — grounded in 2026 human-AI teaming/delegation research (Google DeepMind's Intelligent AI Delegation, the Interposition Problem, HAIF, the Delegated-Autonomy Boundary paper, medical-AI oversight conditions…

1 1mo ago A 113 tokens

VCVinh/Awesome-Research-Development

Skill Claude CodeCodex

How to actually make multi-agent debate/council/board deliberation improve decisions, grounded in 2026 research — not just running N perspectives and assuming more voices means better answers. Covers why vanilla multi-agent debate often underperforms simple majority vote, which decision protocol (voting vs. consensus)…

1 1mo ago A 132 tokens

VCVinh/Awesome-Research-Development

Skill Claude CodeCodex

General-purpose infrastructure for any autonomous, large-scale research pipeline — not specific to one project. Use whenever building, running, or trusting a pipeline that searches many sources (dozens to thousands), digests them, deeply comprehends them (not just restates them), and produces the widest plausible set…

1 1mo ago A 0 tokens

VCVinh/Awesome-Research-Development

Skill Claude CodeCodex

Trending, real (arXiv-grounded, 2025-2026) techniques for continual/lifelong learning, agent harness engineering, and recursive self-improvement (RSI). Use when designing an agent that must keep getting better across sessions without catastrophic forgetting, when optimizing the harness…

1 1mo ago A 158 tokens

VCVinh/Awesome-Research-Development

Skill Claude CodeCodex

When and how to put a semantic layer (Cube, dbt MetricFlow, or bespoke) between an LLM agent and a data warehouse — governed metrics/dimensions/access-rules instead of letting an agent invent raw SQL. Use when an agent needs to query financial/business data (revenue, positions, user metrics) and the same question must…

1 1mo ago A 130 tokens