Lab Skills
Ten ShareAI Lab skills: source-level architecture research, review of past agent sessions, meeting coaching and article adaptation
- 316 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- `review-ai-conversations` reads the local conversation stores of Claude Code, Codex, opencode, Grok Build and Cursor, which may hold secrets and confidential data
- `agent-builder` includes an `init_agent.py` script that creates an agent project on the Anthropic Python SDK
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/lab-skillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- `review-ai-conversations` reads the local conversation stores of Claude Code, Codex, opencode, Grok Build and Cursor, which may hold secrets and confidential data
- `agent-builder` includes an `init_agent.py` script that creates an agent project on the Anthropic Python SDK
This is third-party code. Review the repository files before installing.
What it does
Lab Skills is a collection of ten skills that the ShareAI Lab team distilled from its own work. The skills follow the Agent Skills specification and are not tied to one product. Research: `deep-architecture-research` first clarifies the question, then studies architecture through source, history, documentation and community feedback, `neural-mechanism-research` explains how neural architectures work through computation, geometry and gradients, and `understanding-first-report` assembles long-running work into a clear report. `review-ai-conversations` pulls local conversations from Claude Code, Codex, opencode, Grok Build and Cursor, then searches, compresses and compares them. `vibe-coding` turns the agent into a disciplined developer with transparent decisions and proportionate verification. For meetings there are `meeting-coach-leader` and `meeting-coach-worker`: a leader prepares and reviews a meeting, a worker works out expectations and prepares an update. `agent-builder` helps design AI agents and ships starter code on the Anthropic Python SDK. `skill-judge` scores a skill package against a rubric. `media-writer` rewrites a technical idea for WeChat, Hacker News, Reddit, Medium, X, Dev.to or LinkedIn in that platform's tone.
Who it is for. Engineers and technical leads who study other people's systems, review past work with agents and prepare publications and meetings.
Good fit when
- You pick a framework or runtime for agents and compare options by source rather than by descriptions
- Dozens of Claude Code and Codex sessions piled up over a week and you want a map of questions and open threads
- Long work with an agent spans many turns and the outcome has to be reported to a manager
- A skill is checked with the `skill-judge` rubric before publication: does it trigger on time and add knowledge
- A technical note has to be rewritten for Hacker News and for WeChat in different tones
Not a fit when
- You need one narrow skill for a specific stack: the collection covers research, meetings and publishing, with no individual frameworks in it
- Agent conversations hold confidential data and the `review-ai-conversations` skill must not read them
- You need plugins and general tutorials: the repository holds skills only
- Skill descriptions are in English and the team works in other languages: adapt them to your own
Example request
Use deep-architecture-research and compare from source how three agent runtimes store sessions and memoryLimitations
- The collection is derived from one lab's practice and does not claim to be a universal standard. Check and change its assumptions about your tools and organization.
- `review-ai-conversations` reads the local conversation stores of Claude Code, Codex, opencode, Grok Build and Cursor, and its adapters are tied to their formats.
- The `agent-builder` starter code is written on the Anthropic Python SDK, while the architecture guidance is provider-neutral.
- `deep-architecture-research` requires confirming the research framing before it starts.
How to disable. `npx skills remove <name>` removes a single skill, for example `npx skills remove vibe-coding`. `npx skills update` refreshes the installed skills.
FAQ
How do I install Lab Skills in Claude Code?
Run `npx skills add shareai-lab/lab-skills --skill '*' -y` and all ten skills install for the detected agents. A single skill goes in with `npx skills add shareai-lab/lab-skills --skill deep-architecture-research`, and `--list` previews without installing.
What does review-ai-conversations do?
It pulls full conversation text from the local stores of Claude Code, Codex, opencode, Grok Build and Cursor and does what you asked: search, summary, extraction or comparison. Deep analysis with a question map runs only on request.
How does deep-architecture-research differ from ordinary code search?
It first asks you to confirm the research framing: objects, questions, comparison axes and sources. It then studies source, change history, documentation and the community, and backs conclusions with references.
Can I get a skill's prompt without installing it?
Yes, `npx skills use shareai-lab/lab-skills@understanding-first-report` generates a prompt for one skill and installs nothing. That is a way to try a skill on a task before installing.
Do the skills work outside Claude Code?
Yes, the collection targets agents that support the Agent Skills specification. The core instructions are not tied to one product. The `neural-mechanism-research` and `understanding-first-report` packages carry metadata for Codex.
Discussion
No questions or reviews yet.
Related
Skillscafe3310's personal skill collection: spec-driven project work, reviews, reports, browser, 3D and video in one repository
A personal library of 25 skills: posters and storyboards, Google-style code review, lab report reading and scenario simulations
Five skills from one author: arXiv reports, technical diagrams, a report site, a design system library and agent guardrails
A library of 229 skills for work and life: sales, marketing, finance, engineering, design and parallel agent runs
A personal set of six skills: goal briefs for an agent, disk cleanup, AI news digests, project tidy-up, deep research and long-form writing
A set of 13 SKILL.md skills: research, a product doc, decks, articles and audits for any compatible agent
README in short
The README declares the repository a skills-only collection and gives commands to install, preview, update and handle renames. The skill catalog is split into research, engineering, team collaboration, agent development and publishing, each with scenarios and an example request.
README badge
Original SKILL.md
---
name: deep-architecture-research
description: Deeply research technical architecture, source code, mechanisms, SDKs, frameworks, project comparisons, and system-design options across repositories, history, official docs, issues, discussions, and high-quality community signals. Use when the user asks for deep research, source-level understanding, architecture comparison, mechanism satisfaction, or evidence for designing a system. Always map the question and obtain confirmation before beginning substantive research; skip for simple factual lookups or straightforward implementation tasks.
---
# Deep Architecture Research
Outsource the legwork, not the problem definition. Work as a senior engineer reporting to a human architect.
## 1. Confirmation gate
Use one confirmation gate per research objective. A follow-up that stays inside an already confirmed scope continues without restarting the gate.
Before searching, cloning, or delegating substantive research, send a 500–1,000 Chinese-character framing brief in the user's language. Keep it near 500 characters when possible and include one compact box diagram.
The brief must make four things explicit:
1. **Objects**: what systems, projects, versions, or relationships will be studied.
2. **Questions**: the decisions, mechanisms, and satisfaction gaps to resolve.
3. **Dimensions**: orthogonal axes, layers, dependencies, boundaries, and scenarios.
4. **Sources and output**: repositories, history, docs, community signals, and intended artifact.
Use this shape, adapting labels to the task:
```text
┌──────────────┐ ┌────────────────┐
│ Research objects│──>│ Core questions │
└──────┬───────┘ └───────┬────────┘
│ │
▼ ▼
┌──────────────┐ ┌────────────────┐
│ Dimensions │────>│ Evidence/output │
└──────────────┘ └────────────────┘
```
End by asking whether the map is correct and what should be added or removed. Stop there. The gate is complete only after the user confirms or corrects the scope.
## 2. Research after confirmation
First, organize work by questions and mechanisms, not by vendor names. Define bounded work packages and the evidence each must return.
**MANDATORY — READ ENTIRE FILE:** Before gathering evidence, read [references/evidence-protocol.md](references/evidence-protocol.md). It defines source priority, local-clone practice, history windows, community use, evidence labels, and code-example rules.
Delegate independent work when it improves coverage or cross-checking. Each agent receives one bounded question and returns: