Missing toolchain vs. a real failure
Why "npm not found" is reported as an environment problem, not a code failure the agent tries to fix.
A check can fail for two very different reasons: your code is actually broken, or the tool the
check needs (npm, a specific compiler, a linter) isn't installed where the check runs. Reqursor
Development treats these differently on purpose.
What happens
A result like npm: command not found is classified as an environment problem, not a code
failure. That distinction matters for what happens next:
- It's reported plainly, as a missing tool, not as a failing test to investigate.
- The agent never tries to "fix" it by editing your code or deleting the check that needs it. A missing toolchain doesn't get patched around. It gets flagged for you.
- It skips straight to needing your attention rather than burning a remediation attempt on something no code change can solve.
Where you'll actually see this
Ideally, before it ever affects a ticket: step 5 of the setup wizard runs Run checks on base branch and calls out anything missing right there. You can re-run the same check from Project → Check base branch any time, for example after changing your Docker image or adding a new check.
Fixing it
- On Docker, extend the base image with the tool your project needs. See Adding your project's toolchain.
- On bare metal, install the tool on the host the worker runs on.
Once the tool is in place, re-run Check base branch to confirm, and any ticket that was blocked on it can run normally.