Work with HornDB's nightly LDBC SPB-256 benchmark — find runs, grab the cumulative sqlite artifact, and query the trend series (editorial-qps, aggregation-qps, duration-s) for HornDB vs GraphDB vs Oxigraph. Use when investigating nightly benchmark numbers, a QPS regression or A/B gap, the harness.sqlite trend DB, or the spb-256 nightly job.
Use for any question about a codebase, its architecture, file relationships, or project content — especially when graphify-out/ exists, where the question should be treated as a graphify query first. Turns any input (code, docs, papers, images, videos) into a persistent knowledge graph with god nodes, community detection, and query/path/explain tools.
Diagnose HornDB performance via its Prometheus metrics with a symptom→metric decision tree. Use when a HornDB query, materialization, or data load is slow or regressed, when investigating latency/throughput, or when asked which metric explains a performance issue.
This skill should be used when writing, reviewing, or optimizing SIMD-vectorized database/columnar kernels in Rust — "vectorize this kernel", "SIMD in Rust", "autovectorization", "std::simd", "core::arch intrinsics", "target_feature", "multiversion", "AVX2/AVX-512/NEON", "columnar scan/filter/take", "bitpacking decode", "avoid gather/scatter", or "scalar fallback parity test". Covers the layered strategy (autovec → portable SIMD → intrinsics), runtime dispatch, maintainability, and memory layout for arrow-rs/datafusion/polars/vortex-style code.
This skill should be used when the user wants to add, edit, test, or debug an OWL 2 RL rule in crates/owlrl/rules.toml, add a vocabulary term to crates/owlrl/src/vocab.rs, inspect the compiled output of a rule, or understand how the owlrl codegen pipeline works. Triggers include phrases like "add a rule", "new OWL rule", "edit rules.toml", "add a vocab term", "rdf:type / rdfs:* / owl:* term", "delegate to closure", "show the compiled rule", "inspect generated_rules.rs", "fire_<id>".