lezhang7/autoresearch-valhalla (valhalla)

Use BEFORE committing to an AI/ML research direction, experimental design, training-run config, or autoresearch agent's next step — and when debugging a failing run (loss spikes/NaNs, eval that looks too good, RL reward hacking, RAG/retrieval quality, fine-tuning regressions, distributed-training hangs). Searches Valhalla, a shared knowledge base of others' AI-research experiences (failures and successes with root cause + lesson), so you avoid repeating known dead ends. Also use to record a new experience after a run resolves. Triggers on "design this experiment", "what learning rate/batch size", "why is my training diverging", "plan the research", "ablation", "is this eval contaminated", "should we try X approach", autoresearch / agentic-research loops.

Claude Skills are markdown + scripts loaded into Claude Code; scanned with the same pipeline.

Source:

github.com/lezhang7/autoresearch-valhalla

Provenance:

Found by crawling (provenance, not a security rating)

We found this by crawling public sources and nobody has attested to it. That is how most of the registry gets here, and it is not a mark against the artifact — it only means no authenticated act of publication is on record. An artifact can be the vendor's own and still appear here.