richard-gyiko

Select optimal LLM(s) for a task based on skill requirements, budget, and constraints. Uses the `which-llm` CLI to query benchmark data from Artificial Analysis and capability data from models.dev.

Transform and export data using DuckDB SQL. Read CSV/Parquet/JSON/Excel/databases, apply SQL transformations (joins, aggregations, PIVOT/UNPIVOT, sampling), and optionally write results to files. Use when the user wants to: (1) Clean, filter, or transform data, (2) Join multiple data sources, (3) Convert between formats (CSV→Parquet, etc.), (4) Create partitioned datasets, (5) Sample large datasets, (6) Export query results. Prefer this over in-context reasoning for datasets with thousands of rows or complex transformations.

Use this skill when a task requires operational access to Redmine through the local `rdm` CLI, such as checking connectivity, reviewing or updating issues, logging or analyzing time entries, resolving project or user IDs, or working with Redmine custom fields, even if the user does not explicitly mention Redmine or the CLI.

Minimal spec-driven documentation model for a repo's `docs/` tree — capability specs (with optional product-intent `Why` blocks and executable BDD `.feature` sidecars), ADRs, a `system.md` map, YAML frontmatter for filtering/preload, and an enforcement loop (git hooks → CI → linters) that makes rules self-defending. Use whenever creating, organizing, or maintaining project docs; deciding whether a change needs a doc/spec/ADR update; writing or reviewing capability specs, ADRs, PRDs, or BDD/Gherkin/Cucumber scenarios; adding frontmatter to docs; setting up checks that enforce architecture or documentation rules; reviewing PRs for documentation hygiene; or migrating an existing `docs/` tree into the spec/decision/system layout. Reach for this even when the user says "write up why we built this", "add a feature file", or "enforce our module boundaries" without naming docs explicitly.

Use Mallard to verify local code edits and scope refactor impact with a deterministic structural code index. Trigger after editing code; before renaming, removing, or changing public symbols; and when asked who calls a symbol, what tests exercise it, what changed between SHAs, or whether a refactor missed call sites. Mallard returns citation-grounded symbol records from per-SHA DuckDB indexes for Rust, Python, TypeScript, and JavaScript.