Yang Liu
Community@yang-l · New Zealand
Yang Liu publishes Claude Code skills for Terraform IaC, design-doc editing, LLM coding-pitfall countermeasures, session wrap-ups, and config auditing.
Agent Skills by Yang Liu
Showing 7 vetted skills indexed across 1 GitHub repositories.
yang-lterraform
Guides Terraform and OpenTofu infrastructure-as-code authoring, testing, and security practices.
yang-ldesign-doc
Tightens design docs, RFCs, and tech specs while preserving rationale and structure.
yang-lcoding-pitfalls
Documents countermeasures for nine LLM coding failure modes and agentic leverage patterns.
yang-ldesign-doc-full-pass
Orchestrates parallel read-only agents to edit large design docs section by section.
yang-lwrapup
Ends work sessions with two grounding questions that surface concrete risks and blind spots.
yang-ldistill-skill
Mines past Claude Code session transcripts to draft reusable skills from completed workflows.
yang-lconfig-optimise
Audits and optimises Claude Code config files using a seven-agent research and review workflow.
Frequently Asked Questions About Yang Liu
FAQPage SchemaWhat tasks can I accomplish with Yang Liu's skills?▼
You can write and test Terraform/OpenTofu/Pulumi infrastructure code, tighten design docs and RFCs, apply countermeasures to nine documented LLM coding failure modes, run end-of-session grounding check-ins, distill past sessions into reusable skills, and audit Claude Code config files.
Who are these skills designed for?▼
Platform and infrastructure engineers working with HCL and AWS, tech leads preparing RFCs and architecture specs for review, and developers using Claude Code who want to catch AI-generated code defects, keep configs lean, or convert past sessions into reusable skills.
How do the design-doc skills work in practice?▼
The design-doc skill trims a doc to its one-sentence aim, moves exact values into appendices, and cuts restatement while preserving rationale. The design-doc-full-pass variant fans out one read-only agent per section, collecting verbatim old-to-new string edits applied under a structure-and-anchor guard.
What does the coding-pitfalls skill cover?▼
It provides countermeasures for nine LLM coding failure modes: silent assumptions, abstraction bloat, unscoped side effects, self-referential test blindspots, N+1 batching ignorance, security vulnerability amplification, debugging decay, deprecated API hallucination, and multi-agent coordination failure, plus leverage patterns like tests-first.
Are there prerequisites for using these skills?▼
Skills run inside Claude Code sessions; several specify high-effort models (opus, fable) or inherit the session model. The distill-skill skill requires prior session transcripts to mine and hands off to skill-creator for authoring, while config-optimise requires explicit approval before modifying any file.