
skills
Agent skills for Claude Code, Codex, and skills.sh: agent-rsi retrospective loop, software design, technical writing, coding discipline, PR evidence
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up skills (claude-plugin project) into my current project. Found on https://claudeers.com/skills-33 Repo: https://github.com/luoling8192/skills Homepage/docs: — Detected install method: claude-plugin → /plugin install skills@luoling8192/skills Category: skills. Platforms: cli. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: unknown; community-verified: false. Confirm the source before running anything.
⚠ Unverified / not recently updated — review before pasting a run-this config.
/plugin marketplace add luoling8192/skills /plugin install skills@luoling8192/skills
git clone https://github.com/luoling8192/skills
// compatibility
| Platforms | cli |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | MIT |
| Pricing | open-source |
| Language | Python |
skills
English | 中文
Agent skills by @luoling8192, in one repository. Works with Claude Code, Codex, Cursor, and every other agent skills.sh supports.
Skills
| Skill | Use it for | Source |
|---|---|---|
agent-rsi | Retrospective after every task: counts mistakes and proven methods, writes regression tests, ADRs, and postmortems, and promotes anything seen twice into AGENTS.md | this repository |
software-design-philosophy | Design review based on A Philosophy of Software Design: deep modules, information hiding, complexity | software-design-philosophy-skill |
technical-writing | Chinese internal technical writing: design docs, review notes, postmortems, sharing posts | technical-writing |
ai-coding-discipline | Coding discipline that prevents common AI coding anti-patterns | ai-coding-principles |
ddia-principles | Data system design based on Designing Data-Intensive Applications | ai-coding-principles |
create-pr-with-evidence | Pull requests with verifiable change context, architecture evidence, and before/after screenshots | create-pr-with-evidence-skill |
Install
All skills, for the agents detected on your machine:
npx skills add luoling8192/skills
One skill, globally, for specific agents:
npx skills add luoling8192/skills --skill agent-rsi --global --agent claude-code --agent codex
List without installing:
npx skills add luoling8192/skills --list
Update later with npx skills update.
agent-rsi with its Stop hook (Claude Code)
npx skills add installs skill files only, so the agent runs agent-rsi when it decides to. To have Claude Code ask for a retrospective automatically after every task that edited files, committed, opened a pull request, or deployed, install the plugin, which adds a Stop hook:
/plugin marketplace add luoling8192/skills
/plugin install agent-rsi@luoling8192-skills
The hook needs python3. It asks at most once per stretch of new work. A read-only stretch passes through unless it spent 15 minutes or more in tools and the stats script found a time sink. The prompt quotes the top findings.
For agents without hooks, add a line to your global rules file (~/AGENTS.md, ~/.codex/AGENTS.md, or ~/.claude/CLAUDE.md):
- After any task that changed code, config, pull requests, or deployments, run the `agent-rsi` skill.
How agent-rsi works
- After a task, the agent runs
scripts/session-stats.pyon the session transcript (Claude Code or Codex). The script reports where tool time went: slow command families, calls that are never fast, repeated commands, failures, and polling loops. Then the agent finds the root cause of anything that went wrong or took too long. - It logs the root cause in
agent-docs/ledger.mdat the workspace root, or in~/agent-docs/when every project on the machine needs the lesson. A lesson never lives only in one directory's agent memory. A repeat of the same root cause increments the existing row's count instead of adding a new row. - A failure produces a regression test and, where warranted, an ADR, a postmortem, or a shared lesson in
agent-docs/lessons/. - When a count reaches 2, the lesson becomes a rule in the narrowest
AGENTS.mdthat covers it: a repository's, a workspace's, or your global~/AGENTS.md. - A scheduled sweep runs the script over the last 7 days of sessions, so it also finds costs that are small in each session. A sweep only proposes rules. A person reviews them before they reach a rule file.
Skills are loaded on demand; rules are loaded every session. That is why agent-rsi is a skill, while the rules it writes go into AGENTS.md.
Repository layout
skills/ synced from the source repositories; do not edit here
plugins/agent-rsi/ agent-rsi skill and Claude Code plugin (Stop hook); maintained here
sources.json which source repository and path each synced skill comes from
scripts/sync.py copies the sources into skills/
A workflow syncs skills/ from the source repositories every day, and you can run it by hand from the Actions tab. To change a synced skill, open a pull request against its source repository; pull requests here that touch skills/ fail CI.
To add a skill from another repository, add an entry to sources.json and run python3 scripts/sync.py.
License
MIT. Each synced skill keeps its source repository's LICENSE.
// faq
What is skills?
Agent skills for Claude Code, Codex, and skills.sh: agent-rsi retrospective loop, software design, technical writing, coding discipline, PR evidence. It is open-source on GitHub.
Is skills free to use?
skills is open-source under the MIT license, so it is free to use.
What category does skills belong to?
skills is listed under skills in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/skills-33)
// retro hit counter
[](https://claudeers.com/skills-33)
// reviews
// guestbook
// related in Claude Skills
An agentic skills framework & software development methodology that works.
AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, and more). Turn any folder of code, SQL schemas, R scripts, shell scripts, docs,…
💫 Toolkit to help you get started with Spec-Driven Development