
sanyuan0704/sanyuan-skills
5 skills8k installs18.1k starsGitHub
Install
npx skills add https://github.com/sanyuan0704/sanyuan-skillsSkills in this repo
1Code Review ExpertCode Review Expert is an agent skill that turns your current working-tree or staged git changes into a structured senior-engineer review. Solo and indie builders install it when they want merge-ready feedback without immediately rewriting code—the skill defaults to review-only unless you ask for implementations. It starts with preflight context: git status, diff statistics, and the full patch, then optionally traces related modules with search tools to understand entry points and critical paths like authentication, payments, and data writes. Each issue is classified with a four-level severity model from Critical through Low, with clear guidance on what must block merge versus optional polish. The workflow also accounts for edge cases such as empty diffs, reviewing only staged files, or breaking very large patches into per-feature batches so the agent stays accurate. It fits teams shipping SaaS, APIs, and CLIs who already use git-based workflows with Claude Code, Cursor, or Codex and need consistent architecture and security lenses before CI or human review.2.7kinstalls2SigmaSigma is a personalized one-on-one AI tutor agent skill grounded in Bloom’s 2-Sigma finding—that mastery-oriented tutoring can lift performance far above typical classroom instruction. It is built for solo and indie builders who must repeatedly climb new stacks, APIs, and product domains without a human mentor. Instead of passive reading or generic explanations, Sigma runs a structured tutoring workflow: Socratic prompts, adaptive pacing, misconception diagnosis, spaced repetition on resume, and interleaved checks across topics you already cleared. Progress is explicit: you only move forward when rubric scoring shows you meet the mastery bar (about 80% understanding), which keeps agent-assisted learning honest before you commit code or architecture. Sessions can emit HTML dashboards, Excalidraw concept maps, and generated images so abstract ideas stay concrete. Install it when you want durable competence—not a single summarized paragraph—across Idea research, Validate scoping, Build-time deep dives, or ongoing Operate learning after incidents.2.1kinstalls3Skill ForgeSkill Forge is a meta skill for solo and indie builders who ship with Claude Code and need reusable agent capabilities instead of one-off prompts. It treats each skill as an onboarding guide: procedural steps, domain guardrails, and optional scripts that make outputs more consistent without bloating context. The skill enforces a token budget mindset—challenge every paragraph, default to what the model already knows, and keep SKILL.md under roughly five hundred lines while splitting heavy reference into on-demand files. Use it when you are creating a new skill from scratch, improving an existing SKILL.md, refactoring structure, debugging flaky instructions, or packaging a folder for distribution. It fits the build phase first because you are extending your agent stack, but the same workflow applies when you iterate on skills after code review or production learnings.1.6kinstalls4Book Studybook-study provides markdown templates for a personal or team book wiki organized into concepts, cases, models, and chapter summaries. Solo builders and indie operators use it whenever they ingest non-fiction, technical, or business books and want durable, linkable knowledge instead of scattered highlights. Every template separates source fidelity from your restatement, asks for real-world application, and wires backlinks so ideas compound across reads. Cases explicitly document what a story proves and how trustworthy it is; models spell out steps, limits, and relationships to other frameworks. The skill is methodology, not a scraper: you still read and write, but the agent outputs consistent pages your second brain or coding agent can query during Validate scoping, Build tradeoffs, or Grow positioning—without re-deriving arguments from memory each time.854installs5Wiki Ingestwiki-ingest is an agent skill that gives solo builders repeatable Markdown scaffolds for growing a personal or product wiki from scattered reading. Concept pages distill definitions and context; product pages capture positioning, capabilities, and use cases; pattern pages document problems, solutions, and trade-offs; comparison pages structure A-vs-B tables with a selection guide. Every template enforces Related Pages links and cited Sources so agents do not invent orphan notes. Use it when you are consolidating research into durable docs for yourself, a micro-SaaS, or an agent knowledge base—not when you only need a single README. The excerpt is template-heavy; expect your agent to map incoming material into the closest page type and suggest missing cross-links.720installs