菜单
Use when auditing the ABI Zig codebase, reviewing feature parity, checking stale docs, finding safety risks, or preparing code review findings for GPU, WDBX, MCP, CLI, AI routing, connectors, scheduler, memory, or plugin changes.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when defining ABI framework goals, roadmap, task lists, modernization plans, verification plans, or cross-cutting work across GPU, WDBX, AI model completion/training, MCP, TUI, CLI, and docs.
Use when configuring zls, OpenCode LSP, Zig diagnostics, abi-lsp workflows, editor integration, or troubleshooting Zig 0.17 language-server/build issues in the ABI repo.
Use when redesigning or refactoring the ABI Zig framework, planning module boundaries, splitting features, simplifying APIs, or changing GPU, WDBX, MCP, CLI, AI, scheduler, memory, plugin, or connector architecture.
Use when cleaning, syncing, or configuring hidden AI-tool directories such as .claude, .config/opencode, .opencode, .gemini, .agents, .codex, .copilot, or plugin and skill stores.
Use when writing, reviewing, migrating, or debugging Swift 6.3 code, especially strict concurrency, Sendable, actors, SwiftPM, Swift Testing, macros, and Apple-platform builds.
Use when writing, reviewing, migrating, or debugging Zig 0.17 code, especially build.zig, std.Io, std.process.Init, allocator, import, POSIX, and Darwin linker failures.
| name | abi-codebase-audit |
| description | Use when auditing the ABI Zig codebase, reviewing feature parity, checking stale docs, finding safety risks, or preparing code review findings for GPU, WDBX, MCP, CLI, AI routing, connectors, scheduler, memory, or plugin changes. |
Use this skill for ABI framework audits, reviews, and parity checks. Read tasks/lessons.md and tasks/todo.md first. Treat executable config and source as truth. Inspect mod/stub pairs for public APIs. Prioritize correctness, races, ownership, protocol regressions, and missing tests. Verify with ./build.sh check, ./build.sh check-parity, zig build test-integration, and zig build benchmarks as relevant.
