LinaPro's AI skill system has two categories:

• Externally installed skills: Installed manually with npm/npx commands. Once installed, they apply globally across projects, such as OpenSpec, goframe-v2, and find-skills.
• Project-bundled skills: Stored with the project source under .agents/skills/. They require no extra installation and are loaded automatically when you work in the project directory with AI tools such as Claude Code.

Skills Overview

Skill	Use case	Prerequisites
openspec	Optional spec-driven workflow tool, recommended for this project	Recommended
goframe-v2	GoFrame-specific AI skill that improves backend code generation quality	Recommended
find-skills	AI skill marketplace search tool for choosing suitable skills	Recommended
frontend-design	Create high-quality, distinctive frontend interfaces	None
frontend-patterns	React / frontend development best practices and patterns	None
git-commit-push	Generate a conventional commit message and push to the remote	None
git-worktree	Create an isolated Git worktree for a development task	None
karpathy-guidelines	Behavioral guardrails that reduce AI coding mistakes	None
lina-archive-consolidate	Consolidate archived OpenSpec iterations into unified directories, typically for CI automation	None
lina-auto-archive	Automatically scan and archive completed OpenSpec changes, typically for CI automation	Requires OpenSpec CLI
lina-e2e	E2E test case naming and management standards	Requires Playwright
lina-feedback	Track and fix issues reported after OpenSpec changes	None
lina-perf-audit	Comprehensive backend API performance audit, manually triggered	Requires a running LinaPro service
lina-review	Structured code and specification review	None
openspec-explore	Thinking-partner mode for requirement exploration and problem analysis	None
openspec-propose	Generate a complete OpenSpec change proposal in one step	None
openspec-apply-change	Implement work according to an OpenSpec task list	Requires OpenSpec CLI
openspec-archive-change	Archive completed OpenSpec changes	Requires OpenSpec CLI
playwright-cli	Browser automation and Playwright testing	Requires Playwright

How to Use Skills

Externally installed skills: Follow each skill's installation command. After installation, Claude Code recognizes and applies the skill automatically in matching scenarios.

Project-bundled skills: AI tools automatically load all skills under .agents/skills/ when running in the project directory. To use a skill, describe the scenario directly in the conversation, or invoke it explicitly with a slash command such as /git-commit-push.

Externally Installed Skills

The following three skills require manual installation. After installation, they apply globally across all projects.

openspec

OpenSpec is the recommended command-line tool for LinaPro's spec-driven workflow. After installation, workflow skills such as /opsx:explore, /opsx:propose, /opsx:apply, and /opsx:archive automatically use it as the underlying engine. It is a recommended part of the LinaPro AI development workflow.

When to use: Running the full OpenSpec workflow, including exploration, proposal creation, implementation, and archiving.

Installation:

npm install -g @fission-ai/openspec@latest

After installation, run opsx --version to confirm it is available.

goframe-v2

goframe-v2 is a Claude Code skill tailored for LinaPro backend Go code. It includes GoFrame coding conventions, ORM usage patterns, and best-practice examples. When writing or modifying backend Go files, implementing service interfaces, creating APIs, or working on database operations, the skill activates automatically and improves alignment with framework conventions through code generation, error diagnosis, and performance optimization guidance.

When to use: Developing lina-core backend code, implementing business interfaces, or writing data access layers.

Installation:

npx skills add github.com/gogf/skills -g

find-skills

find-skills is an AI skill marketplace search tool that helps developers quickly discover and evaluate skills suited to the current project. When you need to introduce new AI assistance capabilities but are unsure whether an existing skill is available, find-skills makes skill selection more efficient.

When to use: Introducing new AI skills into the project or evaluating the current skill ecosystem.

Installation:

npx skills add vercel-labs/skills --skill find-skills -g

Project-Bundled Skills

The following skills are maintained with the project source under .agents/skills/. They require no installation and load automatically when an AI tool is used in the project directory.

frontend-design

High-quality frontend interface design guidance. This skill helps AI avoid generic interface output and produce frontend pages with a recognizable visual direction and stronger design quality.

It covers design intent analysis, including page purpose, tone, and differentiation; typography and color selection; motion and spatial composition; background and visual detail treatment; and the definition and precise execution of interface concepts.

When to use: Building a new page, redesigning an existing interface, or needing visual style guidance for frontend work.

How to invoke: Describe the interface requirement in the conversation and the AI applies the design guidance automatically, or invoke /frontend-design explicitly.

frontend-patterns

Best practices and design patterns for React and modern frontend development. This skill gives the AI guidance for component composition, custom Hooks, state management, performance optimization, form handling, error boundaries, motion, and related frontend concerns.

When to use: Building frontend components, choosing a state management approach, implementing performance optimizations, or improving accessibility.

How to invoke: Activates automatically during frontend development work, or invoke /frontend-patterns explicitly.

git-commit-push

Reviews the current Git workspace changes, generates a repository-compliant commit message, commits all changes to the current branch, and pushes the branch to origin.

Commit messages follow the conventional commit format, supporting type prefixes such as fix, feat, build, and docs, and the summary is derived from the actual diff.

When to use: After completing feature development or a fix and needing to commit and push code.

How to invoke: Tell the AI to "commit the code" or "commit and push", or invoke /git-commit-push.

git-worktree

Creates an isolated Git worktree so the AI can continue work in a separate directory without affecting the state of the current checkout.

This is especially useful for handling multiple feature branches in parallel or exploring a new approach without disturbing the main working directory.

When to use: Isolating a task-specific development environment or working on multiple branches in parallel.

How to invoke: Invoke /git-worktree; the AI derives a task identifier for the worktree branch name.

karpathy-guidelines

A set of behavioral guardrails for reducing common AI coding mistakes, inspired by Andrej Karpathy's observations about LLM coding behavior.

The four principles are: think before coding (state assumptions and tradeoffs explicitly), simplicity first (solve the problem with minimal code), surgical changes (change only what is necessary), and goal-driven execution (define verifiable completion criteria and iterate until they pass).

When to use: All coding tasks, as a baseline constraint for AI-assisted development.

How to invoke: Applied automatically as a baseline guardrail; explicit invocation is rarely needed.

lina-archive-consolidate

Consolidates multiple archived OpenSpec iterations under openspec/changes/archive/ into unified archive directories grouped by functional responsibility.

During semantic consolidation, it preserves the design decisions and implementation details from every iteration while compressing multiple iterations into one cohesive archive, reducing historical directory noise. When the archive count reaches 8 or more, it automatically enters the full OpenSpec workflow: exploration, proposal, and implementation.

When to use: The archive directory has accumulated many scattered iteration records that should be periodically merged.

How to invoke: Invoke /lina-archive-consolidate.

lina-auto-archive

Automatically scans openspec/changes/, batch-archives active changes that are complete, and summarizes the result.

The skill strictly checks each change's completion status: both the OpenSpec status and every task must be complete before archiving is allowed. Incomplete changes are skipped rather than forced through, and the final report lists successful archives and reasons for skipped changes in Chinese.

When to use: Batch-cleaning completed changes or periodically tidying the active change directory.

Prerequisite: Install OpenSpec CLI with npm install -g @fission-ai/openspec@latest.

How to invoke: Invoke /lina-auto-archive.

lina-e2e

Naming conventions, directory structure, and management standards for Playwright E2E test cases. This skill defines test case ID allocation (TC{NNNN}, globally unique and never reused), module-based organization, test file templates, independence requirements, and page object model usage.

When to use: Writing, maintaining, or reviewing E2E test cases.

Prerequisites: Install Playwright and test dependencies:

cd hack/tests
pnpm install
npx playwright install --with-deps chromium

How to invoke: The AI applies these standards automatically when writing E2E tests, or invoke /lina-e2e.

lina-feedback

A structured issue tracking and remediation workflow for bugs, missing functionality, experience problems, and test gaps reported after an OpenSpec change has been implemented.

The workflow covers target change identification, issue classification (bug / missing / UX / test gap), incremental spec updates, task list organization, iterative fixes, regression analysis, and comprehensive verification before completion.

When to use: Feedback is reported after an OpenSpec change and needs to be handled systematically.

How to invoke: Invoke /lina-feedback and describe the feedback to address.

lina-perf-audit

A comprehensive LinaPro backend API performance audit workflow, manual trigger only. It audits for N+1 queries, missing indexes, unbounded list responses, duplicate reads, mergeable SQL calls, blocking operations inside loops, and write SQL executed from read endpoints.

The audit runs in three phases: phase 0, environment preparation with database reset, service restart, plugin installation, and load-test data seeding; phase 1, concurrent sub-agent audits for each module; phase 2, report aggregation and persistent issue card generation.

Note

This skill resets the database, restarts services, can take a long time from tens of minutes to hours, and consumes significant Token budget. Use it only when a full performance audit is explicitly needed.

When to use: Explicitly requested systematic performance auditing for all backend APIs.

Prerequisite: A running LinaPro development environment started with make dev.

How to invoke: Say "run lina-perf-audit" or "run a full backend API performance audit" in the conversation.

lina-review

A structured code and specification review workflow triggered after OpenSpec implementation or feedback work is complete and before archiving, ensuring code quality and spec consistency.

Review scope includes backend code review (GoFrame conventions, API i18n compliance, distributed cache consistency), RESTful API review, SQL review, E2E test review, and final report generation. AGENTS.md is the single source of truth for all review standards.

When to use: After an OpenSpec change is implemented and before it is archived as a quality gate.

How to invoke: Invoke /lina-review; the AI determines the review scope from the change.

openspec-explore

Explore mode is an AI thinking-partner mode for requirement analysis and problem investigation. It helps clarify the problem space, investigate the codebase, compare approaches, visualize architecture, and identify risks and unknowns with a curious, non-presumptive approach.

Explore mode focuses on thinking and investigation rather than implementation: it may read files, inspect the codebase, visualize with ASCII diagrams, and produce thinking artifacts such as OpenSpec proposals, but it does not directly implement production functionality.

When to use: Before starting a complex task, or when requirements or technical direction need clarification.

How to invoke: Invoke /openspec-explore and describe the problem or direction to explore.

openspec-propose

Generates a complete OpenSpec change proposal in one step, automatically creating the change directory with proposal.md, design.md, tasks.md, and specs/.

The AI generates artifacts in dependency order so the design document and task list remain internally consistent, avoiding manual assembly of the full change file set.

When to use: Creating a full OpenSpec proposal for a new feature or change.

How to invoke: Invoke /openspec-propose and describe the change you want to implement.

openspec-apply-change

Implements code, documentation, and test changes according to tasks.md in an OpenSpec change directory. This skill is usually triggered after the proposal and design are complete, and is responsible for moving the change from specification to runnable implementation.

When to use: An openspec/changes/<change-id>/ directory already exists and you need to start or continue implementation.

Prerequisite: OpenSpec CLI is recommended.

How to invoke: Invoke /openspec-apply-change, or tell the AI to continue implementing the current OpenSpec change.

openspec-archive-change

Archives a completed OpenSpec change by merging incremental specifications into baseline specifications and moving the change directory into the archive area.

When to use: The change implementation, review, and verification are complete and should be consolidated into the long-term baseline.

Prerequisite: OpenSpec CLI is recommended.

How to invoke: Invoke /openspec-archive-change, or ask the AI to archive the completed change.

playwright-cli

A Playwright browser automation CLI command set covering navigation, clicks, typing, form submission, multi-tab management, Cookie / localStorage / sessionStorage operations, network route interception, DevTools, and other browser automation workflows.

When to use: Automating browser interactions, writing tests, or debugging Playwright tests.

Prerequisites: Install Playwright and the required browsers:

cd hack/tests
pnpm install
npx playwright install --with-deps chromium

How to invoke: Describe the browser operation you need, or invoke /playwright-cli.