| 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"] |
Context Lens
- Run
node "${CLAUDE_PLUGIN_ROOT}/scripts/run-artilens.mjs" lens --latest --data .claude/artilens/lens.data.json.
- Relay the CLI's summary line to the user verbatim (health score, context %, recommendation).
- Start a background subagent with the Agent tool:
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."
- Continue your own task. When the subagent finishes, relay the artifact URL to the user.
- If the user pastes back copy-as-prompt JSON, apply
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.
