Coding & Development

Split a task into the smallest checkable units, tag each with a type and speed class and the clause of the ask it serves, build parallel waves from real data dependencies, gate big maps on an advisor review, and print a plan table for model and machine picks. Also validates simple step plans, write-path overlap and reusable task recipes.

Catch bugs in code that was just written, before it is committed. Run this before every `git commit` that contains code you or the user just wrote — "commit this", "commit my changes", "let's commit", "wrap this up and commit" all mean run this first. Also use when the user says "bug check this", "prevent bugs", or "find bugs before I commit". If another skill makes your commits and runs this one before each of them, let it. Finds candidate bugs in the staged change, proves each one with a regression test that actually fails, refutes the weak ones, fixes what survives, and re-verifies. Docs, specs, tests and work-in-progress commits skip the hunt, not this skill — they still commit through its script, carrying a `Bug-hunter: skipped` trailer. When a caller's brief says "caller commits", run in full but do not commit: leave the change staged and return the trailer lines and any open decisions for the caller to use. Do not use for reviewing someone else's PR or reviewer comments — use a code review for that.

Use when choosing or touching lakehouse storage - Postgres vs warehouse vs lakehouse, Delta vs Iceberg vs DuckLake, slow queries from small parquet files, partitioning, OPTIMIZE/VACUUM/expire_snapshots, MERGE or schema-evolution errors, catalog choice, PySpark collect()/UDF/shuffle/skew, or testing tables locally with DuckDB or polars.

Take an app idea from first description through market research, product definition, Figma design, architecture, planning, autonomous implementation, QA and go-to-market, with persisted state, specialist subagents and approval gates. Use when the user describes a new app idea to evaluate or build, asks to start, continue, or check the status of an App Factory project, or types /app-factory.

Use when designing or reviewing an analytical model (star schema, marts, fact/dimension tables), when metrics double-count or inflate after a join, when choosing surrogate vs natural keys or SCD type 1 vs 2, when deciding raw/staging/intermediate/mart layering, naming, or which tests a model needs, and for PostgreSQL OLTP schema review (normalization, FK indexes, multi-tenant keys).

Use when a new CSV, parquet or xlsx needs profiling before touching it, or when cleaning, deduping or normalizing records, coercing dates/currency/IDs, or reconciling a source of truth against a derived file (OCR'd or scanned documents vs a system export, an upload vs the DB). Covers row-count mismatches, lost leading zeros, duplicates, control totals, anti-joins, exceptions reports.

Turn a rough software idea into approved, implementation-ready specifications. Inspect the repository, ask one highest-priority clarification at a time, propose cohesive feature splits, and write Spec.md and Progress.md only after approval. Use before coding a new feature or redesign.

Implement exactly one approved feature specification with pragmatic test-driven development, strict repository quality gates, sequential verifiable work units, safe Git commits, progress tracking and reliable resume. Use when asked to build docs/specs/<featureName>/Spec.md.

Verify and finalize a completed specification-based software feature. Reconcile tests, approvals, Git commits and progress evidence; preview merge, push and archive changes; obtain explicit approval; then archive completed docs safely. Use after buildtospecs has completed every work unit.

Standardized end-to-end autonomous engineering pipeline for building, verifying, and releasing distributed applications with zero user interruption. Use when tasked with autonomous delivery, direct specification authoring, Quality Gates with clean-context AI courts, or multi-model routing (Architect, Implementer, Clean-Context Reviewers).

Review a custom lint rule the way an experienced ESLint, typescript-eslint, and Flint maintainer would, proving bugs by mutating the rule and running adversarial test cases. Use this whenever the user asks to review, audit, critique, or sanity-check a lint rule, a rule's fixer, its tests, or its docs, or a PR or branch that adds or changes one, even if they only say "take a look at this rule" or give a PR number in a linter or plugin repo. Don't use it for configuring lint rules in a project or fixing lint errors in application code.

Teach the user a programming topic (default stack .NET / C#) through a persistent learning loop (current knowledge, lesson, assignment, updated state) that resumes across sessions. Run it from a single learning hub; each topic gets its own subfolder, created automatically.

Analyzes a codebase and writes a lean CLAUDE.md (under 150 lines) plus supporting .claude/docs/ reference files using progressive disclosure — architectural_patterns.md, testing_strategy.md, build_and_deploy.md, data_model.md, api_reference.md. Use this skill whenever the user asks to generate, create, write, refresh, or improve a CLAUDE.md, AGENTS.md, project context file, onboarding doc, or "docs so Claude understands this repo"; whenever they say their CLAUDE.md is bloated, stale, or too long; whenever they ask to document a codebase's architecture, patterns, or conventions for an AI agent; and whenever they run /init or ask to "index", "onboard onto", "get up to speed on", or "map out" an unfamiliar repository. Prefer this skill over ad-hoc exploration for any of those requests, even when the user does not name CLAUDE.md explicitly.

Builds ready-to-teach Modern Standard Arabic lessons for non-native learners from a theme and level - fully vocalized vocabulary (mufradat) with plural, gender and verb forms, a dialogue (hiwar), one grammar point (qawa'id), exercises across the four skills (maharat), a separate answer key with rubrics, teacher notes with timing, and an optional slide outline or modul ajar. Use when a teacher asks for Arabic teaching material, e.g. "materi bahasa Arab", "mufradat berharakat", "hiwar tema ...", "latihan/soal bahasa Arab", "kunci jawaban", "modul ajar/RPP bahasa Arab", "bahan ajar madrasah/pesantren", "slide pelajaran bahasa Arab", "Arabic lesson plan", "Arabic worksheet" or "Arabic vocabulary list".

Defines the shared Python toolchain, dependency policy, quality gates, security floor, logging schema, and configuration ownership. Use for any Python code task or when another Python skill requires a central rule. Use technical-writing for any documentation produced during the task.

Guide for building multi-language data analysis pipelines, cross-block variable passing, table formulas with org-sbe, and remote block calls in Org Babel.

Audits, plans and writes App Store Optimization for iOS apps - App Store Connect metadata (name, subtitle, keyword field, localizations), keyword research and winnability tests, screenshots and product page assets, custom product pages and in-app events, ratings strategy, Apple Ads, Meta app campaigns (Advantage+ app campaigns, optimization goals, AEM and SKAdNetwork measurement) and their fit with organic search, the App Tracking Transparency prompt and what consent changes for attribution, competitor research, and measurement of changes. Use when asked to improve App Store visibility or conversion, pick or audit keywords, write or review store metadata, plan or critique screenshots, set up or review Apple Ads or Meta app campaigns, write or review the ATT prompt, reconcile installs across Meta, Apple Ads and App Store Connect, or diagnose why an app is not ranking or not converting.

Comprehensive guide for authoring, structuring, executing, and publishing interactive computational notebooks in Emacs Org Mode using Org Babel.

Guide for literate programming, code extraction (tangling), noweb macro expansion, batch compilation, and configuration management with Org Babel.

Review a connected MCP server's tool design and behavior, then produce a safety-conscious scorecard and prioritized fixes. Use when asked to evaluate or put an MCP server through mcp-sauna.

Analyze ncu (NVIDIA Nsight Compute) profiling output: SOL% bottleneck classification, roofline analysis, occupancy diagnosis, memory hierarchy analysis, warp stall analysis, metric interpretation, and programmatic .ncu-rep report analysis. NOT for kernel writing or code generation, Nsight Systems (nsys), host-side profiling, or system-level profiling.

Create and incrementally update exam-focused Anki decks from a university course folder with lecture slides, exercises, solutions, and old exams. Supports mathematics/CS question types, stable note IDs, diagrams, and .apkg packages.

Interviews the person to set a distinctive visual direction — brand, world, graphic devices, signature move — before UI is built, so the result looks like this product and no other. Use when starting or restyling an app, a landing page, a deck or a page that should not look generic, or when a design reads as template, theme or AI default.

Comprehensive SEO analysis by @samknows.ai. Full site audits, single-page analysis, technical SEO (crawlability, indexability, Core Web Vitals with INP), schema markup, content quality (E-E-A-T), image optimization, sitemap analysis, and GEO for AI Overviews/ChatGPT/Perplexity. Triggers on: SEO, audit, schema, Core Web Vitals, sitemap, E-E-A-T, AI Overviews, GEO, technical SEO, content quality.