菜单
Visualize the current Claude Code session's context health, token pressure, hot files, and compact/clear/handoff recommendation. Use when the user asks to see session context, context window, context usage, session health, handoff readiness, or whether to compact.
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
Visualize subagent, Agent tool, and delegated task activity from the current Claude Code session. Use when work was delegated or the user asks what agents/subagents did, what ran in the background, or how tasks were coordinated.
List, find, inspect, diff, resume, or update this project's locally registered artifacts. Use when the user references a previous artifact, asks what artifacts exist, wants artifact history, or wants to continue from an artifact.
Create a draggable planning, kanban, roadmap, or backlog board artifact with copy-as-prompt feedback. Use for task ordering, project boards, sprint planning, prioritization, or moving cards between columns.
Catalog available slash commands, ArtiLens skills, plugins, and contextual recommendations. Use when the user asks what commands/tools/skills are available, what ArtiLens can do, or what to run next.
Create an option comparison artifact with tradeoffs and a recommendation. Use when the user wants to compare approaches, architectures, libraries, plans, implementation choices, pros/cons, or decision options.
Inventory project context files — CLAUDE.md, AGENTS.md, rules, SKILL.md files, plans, and docs — with read-frequency from the current transcript. Use when the user asks what context is loaded, which files shaped the session, or before auditing stale or unused context files.
| name | lens |
| description | Visualize the current Claude Code session's context health, token pressure, hot files, and compact/clear/handoff recommendation. Use when the user asks to see session context, context window, context usage, session health, handoff readiness, or whether to compact. |
| allowed-tools | ["Bash","Read","Agent"] |
node "${CLAUDE_PLUGIN_ROOT}/scripts/run-artilens.mjs" lens --latest --data .claude/artilens/lens.data.json.subagent_type: general-purpose, model: sonnet, run_in_background: true. Prompt template:
"Read <absolute path to lens.data.json> and ${CLAUDE_PLUGIN_ROOT}/references/artifact-authoring.md. Load the artifact-design skill, then author and publish an artifact with the Artifact tool.
Page plan: health score + context % + one-line recommendation (compact/clear/continue) up top; the 10x10 context-window grid with a category legend; token-breakdown; tool-usage and hot-files tables; the handoff block; copy-as-prompt.
All visual design decisions are yours."changes and republish to the same URL (include the prior artifact URL in the subagent prompt).If the recommendation is compact or handoff, read .claude/artilens/lens.data.json yourself and use decision.preservationPrompt in the next /compact, or the handoff block for a fresh session.
Never paste raw transcript content into chat.