← ClaudeAtlas

intellij-healthlisted

Comprehensive health check of IntelliJ IDEA and the Mac it runs on, for a setup where many Claude Code sessions work alongside the IDE. Evaluates IntelliJ's logs (idea.log errors, plugin blame, BSP/sbt trouble, freezes, indexing, MCP tool traffic) and memory usage (IDE heap and footprint, sbt servers and BSP clients per checkout, Kotlin/Gradle daemons, MCP servers duplicated per Claude session, swap), then reports what is wrong and what to do about it. Use this whenever the user asks how IntelliJ or the machine is doing, why things are slow, what is eating memory, whether a change (BSP, shared sbt servers, killing a daemon, disabling MCP servers) helped, wants the IDE log reviewed, or has just rebooted or restored sessions and wants an assessment, even if they don't say "health check".
WamBamBoozle/AgenticIntelliJ · ★ 0 · AI & Automation · score 72
Install: claude install-skill WamBamBoozle/AgenticIntelliJ
# IntelliJ health check Two questions, answered with evidence: **what is IntelliJ's log saying**, and **where is the memory going**. The machine typically runs one IntelliJ with several project windows, a sbt server per Scala checkout shared over BSP, and many Claude Code sessions, each of which starts its own copies of stdio MCP servers. Most memory problems in that setup are duplication (several JVMs or MCP servers doing one job) or leftovers (daemons nobody needs any more), and most log problems are a handful of loud, repeating failures hiding among tens of thousands of harmless warnings. ## Procedure 1. **Collect.** Run both scripts (in parallel; each takes a few seconds): ```bash python3 <skill-dir>/scripts/memory_snapshot.py # add --json for machine-readable output python3 <skill-dir>/scripts/log_scan.py --rotated 1 # current + previous idea.log ``` `log_scan.py` options: `--since "YYYY-MM-DD HH:MM"` to look only at the current incident or the time since a change; `--rotated N` to go further back; `--log-dir` if the IDE's log isn't found automatically; `--top N` for more rows. Both scripts need only the Python standard library; the snapshot is macOS-only. 2. **Dig into what stands out.** The scripts rank and count; they don't explain. For each serious item, look at the evidence directly before drawing a conclusion: - A loud log entry: `grep -n` it in idea.log and read the lines around the first and last occurrence (stack trace, what