pascalwhoop

Create or externally verify YAML tax configurations for any country/year in the universal salary calculator. Use when: - Creating a new country configuration - Adding a new tax year for existing country - Creating variant configs (expat regimes, special tax rules) - Externally verifying existing configs (recommended for all configs) - Fixing bugs in existing configs Triggers: "add country", "create config", "verify country", "fix [country]", "new tax year", "add variant", "30% ruling", "verify test vectors", etc. CRITICAL GUARDRAILS: 1. Test vectors MUST come from official government calculators - NEVER from the engine itself 2. All configs MUST pass `npm run test:configs` before completion 3. Progressive tax systems MUST be verified with marginal rate tests 4. Multi-level taxation (federal+state) requires extra scrutiny 5. Social security systems often have hidden complexity (multiple caps, progressive rates)

Developer agent that implements GitHub issues: reads the issue, creates a branch, writes code, runs tests, and opens a pull request. Also handles rework when a reviewer requests changes on an existing PR. Use when: triggered via @claude /developer on a GitHub issue or PR comment, or when asked to implement a specific issue or fix reviewer feedback.

Adversarial code reviewer that reviews pull requests with high standards, provides structured blocking/non-blocking feedback, and either approves or triggers another developer iteration via @claude /developer. Use when: triggered automatically on a PR via the claude-code-review workflow, or when asked to review a pull request.

Autonomous product manager agent that owns the product backlog and drives the product forward. Researches user needs from GitHub issues, synthesizes patterns, proactively creates well-defined issues for gaps in the product, reprioritizes based on user signal, triages and coordinates the backlog using labels, triggers the developer agent on ready issues, and maintains the pm/ folder (vision, roadmap, timestamped update entries). Use when: running as a scheduled GitHub Actions agent to manage issue flow and product direction, or when asked to triage/coordinate the issue backlog, research what users need, or define new features.

Staff engineer / architect agent that performs broad codebase refactoring to reduce complexity, improve maintainability, and consolidate duplication. Uses ESLint complexity analysis as input to identify hot spots. Creates a PR with focused, safe refactors. Use when: triggered by the weekly architect workflow, or when asked to refactor/simplify the codebase, reduce complexity, or improve code structure.

Streamlined bug fixing workflow that creates GitHub issues, writes reproducible tests, finds fixes, and commits changes. Use when a user reports a bug and wants to: (1) Create a GitHub issue, (2) Write a test that reproduces the bug, (3) Find and implement a fix (with user iteration), (4) Commit with the issue number. Triggers: "fix this bug", "there's a bug", "I found an issue", "bug report", "something's broken", etc.

Expert guide for querying the Everycure Knowledge Graph (9.2M nodes, 77M relationships) for drug repurposing, mechanistic pathway discovery, and biomedical research. Use this skill when working with Everycure KG MCP tools to: (1) Find drug-disease connections and repurposing opportunities, (2) Explore mechanistic pathways (drug→protein→gene→disease), (3) Discover gene-disease associations, (4) Avoid path explosion in dense graph queries, (5) Navigate schema quirks and validate queries, (6) Design custom metapaths for novel hypotheses. Critical for biomedical researchers, drug discovery teams, and anyone performing knowledge graph-based drug repurposing analysis.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Initialize a new problem tree for recursive problem solving. Use when the user wants to set up a new problem to solve using the tree decomposition approach.

Analyzes a single problem tree node and either decomposes it into sub-problems OR solves it directly. Never handles multiple nodes or siblings.

Orchestrates end-to-end podcast episode production for drug repurposing stories. Use this skill when the user requests creating a podcast episode for a specific drug-disease pair (e.g., "Create an episode for Aspirin and Cardiovascular Disease Prevention" or "Produce the Metformin PCOS episode"). Automatically runs the full production pipeline: archivist (research) → showrunner (structure) → podcast-writer (scripting) → editor (editorial pass) → speaker (audio generation) → git commit. Handles error recovery with TODO markers and partial work commits.