deniz.in

Markets

Weather

Loading weather

· via dev.to (home feed)

Cline crash diagnostics wrote plaintext API keys to disk every few minutes

Cline's core process for JetBrains crashed with out-of-memory errors roughly every seven minutes, and each ~240MB heap dump captured the user's API key in plaintext, per a GitHub issue.

Cline crash diagnostics wrote plaintext API keys to disk every few minutes

What happened

A Cline user running the JetBrains plugin reported that the tool's standalone cline-core process was crashing with an out-of-memory error roughly every six to seven minutes. According to a dev.to write-up of GitHub issue #15427, filed on 29 October 2026, every crash produced a diagnostic heap snapshot of about 240MB written to ~/.cline/heapsnapshots/ — and those snapshots captured the session's full in-memory state, including the user's provider API key in plaintext.

The snapshots reportedly contained hundreds of near-duplicate copies of that state, so a single key could recur many times inside one file.

Why the process kept crashing

The issue, filed against JetBrains plugin 2.0.9 with core 4.1.26, describes the mechanism: cline-core streams a fully serialized copy of its globalState — chat transcript and API key included — to the JetBrains webview on every state change, with no backpressure on the update queue. During a burst of changes, memory climbed from roughly 270MB at idle to several gigabytes in about 100 seconds before the JavaScript heap gave out and the process exited with code 134.

Maintainer mkondratek confirmed on 2 November 2026 that the crash shares its root cause with an earlier report, issue #15031: the queue never discards stale entries when the IDE falls behind, so pending updates pile up until memory is exhausted.

A separate fix for the key exposure

The same maintainer acknowledged the plaintext-key exposure the reporter flagged and said a dedicated fix for it was in progress, tracked separately from the memory bug. As of 9 November 2026, according to dev.to, only a telemetry commit had landed; neither the OOM fix nor the key-redaction fix had been merged by that date.

What the report does not establish

There is no indication of data loss or external exposure. The heap snapshots and the key stayed on the user's own machine, and the process restarted itself after each crash. In the reported case this is not a breach — it is a credential persisting in a place it was never meant to reach.

What users should check

The dev.to article advises anyone running Cline's JetBrains plugin to inspect ~/.cline/heapsnapshots/ for snapshot files and to rotate any API key that was active while a crash occurred.

Why it matters

Crash diagnostics tend to be treated as inert artifacts — memory captured, attached to a bug report, forgotten. This incident shows why that assumption fails: any tool that holds secrets in memory will leak them into whatever captures that memory, whether heap dumps, core files or error reports. AI coding assistants routinely hold provider API keys, which makes their diagnostics a recurring leak surface. Redaction has to happen at the moment of capture, because nothing downstream will do it automatically.

The episode also illustrates a quieter risk: a diagnostic feature can persist sensitive data at scale without anyone opting in. Hundreds of duplicated 240MB snapshots written every few minutes is simultaneously a disk-space problem and a secret-storage problem — and until both fixes ship, users of the JetBrains plugin are left to discover it on their own.

  • #cline
  • #security
  • #jetbrains
  • #api-keys
  • #ai-coding-tools

Related posts