@lllllllama
Covers science, devops, writing. Works: paper-context-resolver, repo-intake-and-plan, minimal-run-and-audit.
All Skills by @lllllllama
ai-research-explore
by @lllllllama
This skill provides a rigorous and reproducible exploration process for deep learning research. It assists researchers in exploring candidate solutions based on defined tasks, datasets, and benchmarks, ensuring auditable experiments, comparable results, and the generation of high-quality exploratory outputs, thereby accelerating meaningful and innovative research.
analyze-project
by @lllllllama
This skill performs read-only analysis and auditing of deep learning research repositories. It helps users understand model structures, training/inference entry points, and configuration relationships, identifying potential insertion points or suspicious implementation patterns. This provides deep project insights without modifying code or running heavy tasks.
ai-research-reproduction
by @lllllllama
This skill automates the reproduction of deep learning research repositories, ensuring minimal trust, auditable evidence, and scientific rigor. It streamlines the process from reading READMEs, selecting minimal trustworthy targets, coordinating setup, execution, training verification, repository analysis, to generating standardized reports. This significantly enhances the efficiency and scientific credibility of AI research validation, reducing manual intervention and errors.
explore-code
by @lllllllama
This skill serves as a rigorous improvement implementation leaf skill for deep learning research, enabling researchers to perform exploratory code modifications on isolated branches, such as transplanting modules, adapting backbones, or adding adapter layers. It ensures all changes are auditable and rollback-aware, helping researchers conduct low-risk experiments efficiently and systematically record processes and outcomes, thereby enhancing research rigor and traceability.
safe-debug
by @lllllllama
This skill is designed for debugging and auditing deep learning research work. When users encounter issues like tracebacks, terminal errors, CUDA OOM, checkpoint load failures, shape mismatches, or training failures, this Skill provides conservative diagnoses and minimal patch suggestions. It clearly separates debugging fixes from research contributions, helping users quickly identify and resolve problems while ensuring the rigor and efficiency of their research process.
explore-run
by @lllllllama
This skill is designed for executing bounded exploratory runs in deep learning research, such as small-subset validation, short-cycle guess-and-check, batch sweeps, or quick transfer-learning trials. It helps researchers systematically plan and summarize experiments, efficiently evaluating candidate models and methods to accelerate the research process.
run-train
by @lllllllama
This skill executes selected deep learning training commands and standardizes the output of training evidence, ensuring experimental reproducibility. It assists users with startup verification, short-run verification, full training kickoff, or resuming training, automatically recording commands, configurations, seeds, logs, checkpoints, status, and metrics. This significantly enhances the efficiency of deep learning research and experiment management.
paper-context-resolver
by @lllllllama
Resolves key contextual information in AI paper reproduction, supplementing technical details like dataset versions, evaluation protocols, and checkpoint mappings, providing actionable evidence.
repo-intake-and-plan
by @lllllllama
RigorPilot skills empower AI agents to reproduce, improve, and explore deep learning research projects with scientific rigor. It emphasizes meaningful changes, fair comparisons, reproducible evidence, and auditable modifications, driving substantial progress in deep learning research rather than merely chasing scores. The skill focuses on deep learning research repositories, handling environment setup, data, model training, and evaluation.
minimal-run-and-audit
by @lllllllama
RigorPilot Skills are research-first agent skills designed for deep learning experiments. They empower AI agents to reproduce, improve, explore, and audit deep learning research projects with scientific rigor, ensuring meaningful changes, fair comparisons, reproducible evidence, and auditable modifications.
env-and-assets-bootstrap
by @lllllllama
Prepares the runtime environment and dependencies for AI paper reproduction, automatically detecting missing model checkpoints, datasets, and cache directories, supporting various cloud storage.
ai-paper-reproduction
by @lllllllama
Reproduces code repositories from AI papers, automatically analyzing READMEs, identifying training scripts, executing minimal credible runs, and outputting standardized reproduction reports.