Skip to content

Install's skill fetch fails with npm ECOMPROMISED even after its fresh-cache retry #888

Closed

Description

@fpresta0607

The install's skill fetch can fail with npm's ECOMPROMISED ("Lock compromised") even after its retry with a fresh npm cache, and the Install workflow's check then fails the job.

Where

  • Workflow Install, job setup (powershell), step "Check the installed home".
  • Run 38628936476, attempt 1, on pull request feat(voice): dictate with Parakeet TDT 0.6B v2 #882 (feat/voice-accuracy), 2026-10-14T20:16:28Z. The install itself had passed (TestTheSetupInstallsWithDictationReadyOnAFreshRunner, 476.86 s) before this step.

The exact error

the install could not fetch the gh-axi skill and named it in its closing note: npx exited with code 1 after its retry (npm error code ECOMPROMISED), so .claude\skills hold none; a failed fetch, not a broken install

How often

  • Two known: run 37682686124 on main, 2026-10-07 ("Lock compromised", which led to the fresh-cache retry in install.ps1), and run 38628936476 above, after that retry.
  • Of the last 200 Install runs (since 2026-10-09), 15 failed, 36 failed jobs in all, and none of the others shows ECOMPROMISED, so about 1 in 200 runs.

What is known

  • install.ps1 (the skills block after "The tools the fleet drives publish their own skills") runs npx -y skills@<version> add kunchenguid/<skill> ... for gh-axi, chrome-devtools-axi and no-mistakes, and on a failure tries once more with a fresh npm_config_cache of its own. The failure above came after that retry, so a stale entry in the shared npx cache is not the whole cause.
  • ECOMPROMISED is npm's lock library reporting that a lock it holds was updated or went stale under it, which a slow file system (a busy runner, or Defender scanning the cache) can cause.
  • The rerun of the failed job is the workaround today.

Done when

A skill fetch on a runner survives this lock failure, or the workflow tells a failed skill fetch apart from a broken install without failing the job on the former, and the change says which.

Activity

  1. added
    ops-rootA root issue found while the fleet runs, fixed at its cause
    on Oct 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    ops-rootA root issue found while the fleet runs, fixed at its cause

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions