Skill Claude CodeCodex
Validate and plan linear skill chains defined in YAML. Reads IO contracts from skill frontmatter, verifies type compatibility between chained steps, and outputs step-by-step execution plans. Use when the user wants to orchestrate multiple skills in sequence, validate a chain definition, or see an execution plan for a…
Skill Claude CodeCodex
Automatically learn reusable behaviors from agent sessions and evolve them into skills, commands, or agents. Use when the user wants to enable, tune, or review the newer automatic learning workflow, and prefer this over continuous-learning by default.
Skill Claude CodeCodex
Distill the just-finished conversation into a structured Markdown note and optionally save it into the configured L3 knowledge base with automatic frontmatter completion and incremental indexing. Use when the user asks to save or distill the current conversation. Not for browsing historical chat logs.
Skill Claude CodeCodex
Search the local Markdown knowledge base using hybrid (vector + FTS) retrieval. Use when you need to find architecture decisions, past research, configuration guides, or any project-specific knowledge. Supports preset modes for different scenarios (coding, audit, qa, fast). 当用户提到"搜索知识库""查知识""查历史决策""knowledge…
Skill Claude CodeCodex
Capture concise L2 Whiteboard memory entries on top of memory-manager. Use when the user says "记到L2", "写入白板", "提炼成 L2 记忆", "记下这个决策", or wants 1-3 shared decision/action/learning entries extracted from the current task or conversation. Prefer this over calling memory-update.py directly for normal L2 writes.
Skill Claude CodeCodex
Watch Markdown knowledge base directories and auto-trigger incremental RAG indexing on file changes. Use when the user wants to set up automatic L3 index syncing, enable background indexing, or configure filesystem watchers for their knowledge base. 触发词：自动索引、知识库同步、l3 sync、watcher、自动入库。 NOT for manual indexing (use…
Skill Claude CodeCodex
Export SKILL.md frontmatter to MCP (Model Context Protocol) compatible JSON schema. Generates tools/list compatible JSON for any MCP-aware agent to discover and invoke skills. Use when user mentions "MCP", "Model Context Protocol", "导出 MCP", "MCP schema", "tools/list export". 当用户提到"MCP导出""MCP兼容""导出工具列表""MCP JSON"时触发。.
Skill Claude CodeCodex
Manage cross-agent persistent memory using a three-layer model (Identity/Session/Knowledge). Use when you need to search past decisions, actions, or learnings across conversations, or when you need to persist important information for future sessions. 当用户提到"记忆""memory""whiteboard""历史决策""跨会话""保存决策""查找历史"时触发。Prefer this…
Skill Claude CodeCodex
Sync the curated Windows Codex Desktop bridge symlinks from the shared runtime manifest. Use when you want to materialize or validate the Windows Codex skill bridge set against the WSL source of truth. Do not use it for L2/L3 memory writes or OpenClaw-private skills. 当用户提到"bridge sync" "runtime bridge""同步技能""Codex…
Skill Claude CodeCodex
Validate the shared KitClaw runtime contract across WSL, Windows Codex Desktop, shared L2/L3 memory, and OpenClaw bridge links. Use when you need a fast health check for the shared agent runtime or want to detect drift in paths, symlinks, config values, or bridge targets. 当用户提到"runtime doctor" "健康检查""诊断""bridge""path…
Skill Claude CodeCodex
Manage and run scheduled periodic tasks using cron. Supports two levels - Level 1 for agent-free script execution and Level 2 for agent-assisted intelligent tasks. Use when the user wants to automate periodic tasks, install/remove cron jobs, list scheduled tasks, or run a task manually. NOT for one-off task execution…
Skill Claude CodeCodex
A quality gate for skills before they are accepted into the public KitClaw repository. It checks metadata, security, dependencies, compatibility, documentation, and extra files.
Skill Claude CodeCodex
Repository-wide lint workflow for skill frontmatter, description routing quality, and high-risk body inconsistencies. Use when the user wants to lint a shared skills repository, enforce skill metadata conventions, or review skills after a bulk refactor or cleanup.
Skill Claude CodeCodex
Record skill execution logs, generate usage statistics, find unused skills, and produce observability reports. Use when the user wants to track which skills are being used, find unused or high-failure-rate skills, or generate a monthly usage report. NOT for security scanning (use skill-security-audit), NOT for skill…
Skill Claude CodeCodex
Perform static security analysis on AI skill directories — scan for hardcoded credentials, undeclared network requests, permission boundary violations, and IO contract inconsistencies. Use when the user asks to "audit skills", "scan for security issues", or "check skill safety". 当用户提到"安全审计""扫描凭据""检查skill安全"时触发。 Prefer…
Skill Claude CodeCodex
Quality audit workflow for Claude skills and commands, with Quick Scan and Full Stocktake modes. Use when reviewing a skill library for overlap, staleness, or improvement opportunities.
Skill Claude CodeCodex
Sync rules and patterns from conversations to the "Brain" knowledge base. Extracts rules, categorizes them (language, framework, pattern, projects), and updates rules.json and rules.md in the brain directory. Use when the user asks to save a pattern, rule, or best practice to their long-term memory or "Brain".