neozenith

Generate and serve rich HTML companions to markdown discovery documents: paired markdown→HTML rendering (marked + mermaid + data-driven cytoscape/plotly fenced blocks), a vendored draw.io stencil library (~AWS/GCP/Azure/K8s SVG icons) for composing custom architecture diagrams, an injectable design-tokens brandpack, and a reliable localhost server for HTML that pulls pinned CDN libraries. Use when turning a review/architecture/cost markdown doc into an interactive HTML view, when a diagram needs real cloud provider icons, when serving local HTML that file:// breaks, or when building a data-driven graph/chart view. Skip when the deliverable is plain mermaid-in-markdown authoring with no HTML companion, or a Python-rendered architecture diagram image.

Introspect Claude Code sessions - query conversation history, tool usage, event trees, and costs. Use when reviewing past sessions, debugging failures, or recovering user intent post-compaction.

Documentation quality skill with three modes: (1) AUDIT — fan out parallel subagents to corroborate that docs still reflect the reality of the code (commands run, paths exist, signatures match, claims hold) and produce a drift report; (2) WRITE/IMPROVE — author or rewrite docs using the researched lens taxonomy (Diátaxis + distilled style principles from great OSS docs), optionally in the maintainer's personal voice; (3) RESTRUCTURE — reorganize an existing doc, spec, or plan file for structural readability (heading hierarchy, list/table discipline, whitespace rhythm, BLUF) without changing its claims. Use when the user asks to check/audit docs, fix stale docs, write or improve documentation, restructure a doc/spec/plan for readability, or says 'gooddocs'. Add 'voice' to the invocation to write in the maintainer's voice; default is the neutral researched style.

Render and analyze Mermaid.JS diagrams embedded in markdown files. Enforces visual complexity limits AND WCAG color-contrast requirements via automated analysis. Every diagram MUST derive its palette from `resources/color_theming.md` and pass `scripts/mermaid_contrast.ts` before being declared done. Use when creating or updating architecture diagrams, analyzing diagram complexity, improving documentation clarity, or exploring a codebase to produce dual-density architecture diagrams for a README.

Playbook for building and extending project-local developer CLIs and the recurring assets they generate, so the same patterns are never re-derived from scratch. Use when: scaffolding or adding a subcommand to a repo CLI (argparse/Typer, project-root discovery, packaged assets); generating a STATIC HTML single-page viewer (client-side SPA routing, collapsible sidebar, light/dark + rebrandable design-tokens theming, Cytoscape graph or Plotly chart, generate/serve, --inline single-file, --archive zip); templating GitHub Actions workflows from a base template (create/update/validate, trigger-path glob distillation, reusable composite-action assets); using git worktrees to act on another branch/ref safely; autogenerating SVG/PNG diagrams from a stencil pack; or automating sticky PR comments and embedding a CI-run artifact image in a PR comment. Skip for non-CLI app code, or one-off scripts with no reuse. Invoke as /cli or when the user names one of these patterns.

Script-driven gap-analysis planning: a deterministic state machine (pgsm.py) owns the phases of a tiered gap-analysis spec — bootstrap → refinement → validation → decomposition → execution → complete — with evidence-based transition gates, per-state-only instructions composed into a single prompt, durable pause/resume state, and session-log forensics. Use for long-horizon work-breakdown planning where phase tracking must not live in the model's head. Requires uv. Skip for quick single-file plans (use plan-gap) or when no plan document is wanted.

Principled, evidence-based code review of a diff, branch, or PR. Runs parallel perspective passes, validates every candidate finding before reporting, and filters aggressively for precision over volume. Use when the user asks to review code, a diff, a branch, or a pull request with this skill.

Walk a human through any PR or diff, regardless of size. Distills 100s-1000s of changed files into their deeper structure: a few distinct mechanical change patterns (each shape-checked for hidden deviants) plus the genuinely novel changes, then acts as an interactive guide revealing detail incrementally until the reviewer is satisfied. Use when the user says 'walk me through this PR', 'explain this diff/PR', 'what actually changed here', or faces a huge PR they can't review file-by-file. Skip when the user wants a verdict on code quality — that's /codereview.

Runs GGUF chat models and structured extraction (NER, relation extraction, combined NER+RE, summarization) inside SQLite via muninn_chat(), muninn_extract_entities(), muninn_extract_relations(), muninn_extract_ner_re(), muninn_summarize(), and their _batch variants. Supports supervised (labels provided) and unsupervised (open-extraction) modes, and GBNF grammar-constrained JSON output. Use when the user mentions "muninn_chat", "named entity recognition", "NER", "relation extraction", "RE", "LLM in SQL", "GGUF chat model", "Qwen3.5", "Qwen 3.5", "GBNF grammar", "grammar-constrained", "summarization in SQLite", "structured extraction", or wants to run an instruction-tuned model inside SQLite.

Generates text embeddings inside SQLite via muninn_embed() backed by local GGUF models (BERT, MiniLM, nomic-embed-text, BGE-M3, Qwen3-Embedding) through llama.cpp with Metal GPU acceleration on macOS. Covers muninn_embed_model registration, the temp.muninn_models virtual table, composing with hnsw_index for semantic search, and auto-embed TEMP triggers. Use when the user mentions "text embedding", "semantic search", "sentence embedding", "muninn_embed", "GGUF embedding model", "MiniLM", "nomic-embed", "BGE-M3", "Qwen3 embedding", "embed model in SQLite", "text-in semantic-search-out", or wants to embed text directly in SQL.

Runs graph algorithms (BFS, DFS, shortest path, PageRank, connected components, degree / node-betweenness / edge-betweenness / closeness centrality, Leiden community detection) on any SQLite edge table via table-valued functions. Covers the WHERE edge_table = ... constraint calling convention and the persistent CSR adjacency cache (graph_adjacency). Use when the user mentions "graph traversal", "BFS", "DFS", "shortest path", "Dijkstra", "PageRank", "connected components", "centrality", "betweenness", "closeness", "Leiden community", "community detection", "graph algorithm in SQLite", "graph_bfs", "graph_pagerank", "graph_leiden", "CSR adjacency", "graph_adjacency", or wants to analyze a network / dependency / social graph stored in SQLite.

Writes dbt-inspired lineage queries using the graph_select TVF — ancestors, descendants, depth-limited traversal, transitive closures, and set operations on any SQLite edge table. Selector DSL supports +node (ancestors), node+ (descendants), N+node+M (depth limits), @node (closure), space (union), comma (intersection), and "not" (complement). Use when the user mentions "graph_select", "dbt selector", "lineage query", "dependency graph", "upstream", "downstream", "ancestors", "descendants", "transitive closure", "impact analysis", "build closure", "dead code detection", or wants dbt-style node selection syntax on any SQLite edge table.

Builds end-to-end GraphRAG retrieval over a text corpus entirely in SQLite: chunk → embed → extract entities and relations → build knowledge graph → detect communities → label clusters → retrieve via vector seed + graph expansion + centrality ranking. Composes muninn_embed, hnsw_index, muninn_extract_ner_re_batch, graph_leiden, muninn_label_groups, and muninn_extract_er. Use when the user mentions "GraphRAG", "Graph RAG", "knowledge graph RAG", "retrieval-augmented generation", "RAG pipeline in SQLite", "KG retrieval", "entity resolution", "muninn_extract_er", "community labeling", "KG indexing", or wants to build Microsoft GraphRAG- style retrieval on a corpus.

Trains Node2Vec structural graph embeddings (Grover & Leskovec, 2016) from a SQLite edge table and writes them directly into an hnsw_index virtual table. Covers the node2vec_train scalar function, p/q walk bias tuning, window size, negative sampling, epochs, and composing with KNN vector search. Use when the user mentions "node2vec", "graph embedding", "structural embedding", "random walk embedding", "DeepWalk", "node2vec_train", "similar nodes", "graph representation learning", or wants to compute embeddings from graph topology (not text).

Installs, loads, and smoke-tests the muninn SQLite extension across SQLite CLI, C, Python, Node.js, and WASM runtimes. Covers pip install sqlite-muninn, npm install sqlite-muninn, prebuilt .dylib/.so/.dll from GitHub Releases, and from-source builds with make. Use when the user mentions "install muninn", "load muninn", "load the extension", "enable_load_extension", ".load ./muninn", "sqlite3_muninn_init", "sqlite-muninn", "pip install sqlite-muninn", "npm install sqlite-muninn", "muninn.dylib", "muninn WASM", or asks to get started with the library.

Diagnoses muninn build failures, extension-loading errors, platform-specific pitfalls, and runtime issues across SQLite CLI, C, Python, Node.js, and WASM. Covers SQLITE_OMIT_LOAD_EXTENSION on macOS system Python, CMake hangs on Apple Silicon (GGML_NATIVE), Metal GPU toggles (MUNINN_GPU_LAYERS), llama.cpp log verbosity (MUNINN_LOG_LEVEL), ASan-built dylibs failing in Python, GGUF model resolution, and wasm32 vs wasm64 linkage. Use when the user mentions "unable to open shared library", "not authorized", "SQLITE_OMIT_LOAD_EXTENSION", "enable_load_extension fails", "muninn build failed", "CMake hangs", "Segmentation fault muninn", "model not registered", "MUNINN_GPU_LAYERS", "WASM muninn fails", "muninn troubleshooting", or hits a muninn load / build error.

Help a developer decide which data-quality tests to apply to a scope of dbt models, then implement them correctly. Use when someone asks "what tests should this model have?", "how do I test uniqueness / grain / freshness / a foreign key / an enum / a numeric range / SCD2 / an anomaly?", or wants to close test coverage on staging/marts before a PR. Walks the testing-taxonomy decision tree in references/, opens the matching vignette (the worked pattern + why it matters + which dbt package to reach for), and grounds the final implementation in current practice with a web search because vignette code samples and package recommendations date. This is the decide-and-justify guide for specific models — not a whole-project coverage scorer and not a unit-test YAML generator.