searchsim/

AgentArtifactCorpus

Agent configuration artefacts (AGENTS.md, CLAUDE.md, .cursorrules, .mdc rules, system-prompt files) crawled from public GitHub repositories under permissive licences. AAC underpins the compaction, decomposition, and classifier experiments in the CIKM 2026 paper The Compaction Cliff in Long-Running AI Agent Memory.

Code and reproduction: https://github.com/searchsim-org/knowledge-triage Datasheet: see DATASHEET.md in the code repository Access: gated… See the full description on the dataset page: https://huggingface.co/datasets/searchsim/AgentArtifactCorpus.