agent-sdk-builder
Skill: Claude Agent SDK Builder
When the user wants to build an agent with the Claude Agent SDK — whether that's writing their first
query(), wiring up hooks, adding subagents, or deploying to production — follow the steps in this skill to diagnose their intent, surface the right reference, and produce working code.
How to Use This Skill
Trigger: User asks about building, debugging, or extending a Claude Agent SDK agent (Python or TypeScript).
Process:
- Identify the user's goal using the Intent Classifier below.
- Read the reference file(s) listed for that goal — they are in
references/and indexed inINDEX.md. - Produce a working code snippet in the user's language (Python or TypeScript). Default to Python if unspecified.
- Point the user to the exact reference section for deeper reading.
Intent Classifier
More from river-walras/virtus
html-canvas
Generate a rich, self-contained HTML artifact for specs, code reviews, designs, reports, or custom editors. Use when the user asks for an "HTML file/artifact/explainer/report/playground", wants a shareable visual deliverable instead of markdown, or asks to "make a canvas/dashboard/diagram-heavy doc". Trigger phrases include "make an HTML", "html artifact", "html canvas", "html report", "html explainer", "html playground", "visual spec", "html mockup".
6claude-blogs
Answer questions about Claude and Anthropic products using official claude.com/blog posts as a knowledge base. USE when: user asks how Claude Code works, asks about Claude features (MCP, skills, artifacts, prompt caching, tool use, web search, context windows), asks about Claude plans or pricing, asks about Claude integrations (Slack, Microsoft 365, Amazon Bedrock), asks for enterprise AI use cases or case studies, asks about agent skills or multi-agent systems, asks about best practices for prompting or building with Claude. NOT USE when: user is asking about a non-Claude AI product or provider, asking general programming questions unrelated to Claude, asking about Anthropic safety research or model internals not covered in blog posts, or the question can be answered from existing conversation context without needing blog references.
5walrasquant-strategy
Help write, modify, review, and debug WalrasQuant trading strategy files. Use when Codex needs to create or edit strategy scripts under the strategy directory, subclass walrasquant.strategy.Strategy, wire Config and Engine, use subscriptions, event handlers, schedules, indicators, cache access, order APIs, precision helpers, execution algorithms, or wq process commands for WalrasQuant strategies.
2alpha-vault
Build or refresh a portable strategy knowledge base for this repository by bundling notebooks, local markdown notes, and code implementations into a skill-local `ref/` folder. Use when Codex needs to answer research questions about these strategies, locate implementation files, refresh the repository's reference corpus, or search topic summaries from a self-contained skill bundle.
1git-auto-release
Review git changes first, then create an appropriate commit, create or choose a release tag, and push the branch and tag. Use when the user asks to auto commit, tag, push, publish a release tag, or prepare a git-based release after local changes are ready.
1