87/ 100
Low riskLow risk: The observed install activity was limited, but installation failed and the entry phase did not complete. Scanned 23 Oct 2026.
Score
87 / 100Nothing was taken off for what the code did or what running it showed.
What it did
41 events in 73.0sLuna's report
The verdict is Low risk, but the entry phase did not complete and flagged code was not observed running. Pin the scanned version. Most importantly, later entry-point behaviour was not verified.
We installed the repository in a throwaway sandbox and scanned code used by its install process and entry point. Dependencies installed, but the repository’s preinstall script failed, so the entry point did not run to completion.
What it ran
Jev's answers
answered from the evidenceShow all 8 answers
package.jsonscripts/prepare-git-hooks.mjsscripts/lib/package-lifecycle-marker.mjsscripts/runtime-postbuild-shared.mjsShow all 3 answers
Not verified
Behaviour outside the observed install and entry windows
Code paths not triggered by install or the entry point
Show all 7 items
Payloads gated on calendar dates past the accelerated clock window, or on hosts that need real responses
Flagged code that was read but never seen doing anything in the sandbox, so it takes no points: 48 reachable finding location(s), -268 floored at -40. (-5: Code reads the npm login token file at .agents/skills/openclaw-live-updater/scripts/update-main.mjs:88 (install); -5: Code reads the npm login token file at .agents/skills/telegram-e2e-userbot/scripts/published-upgrade-scenario.mjs:304 (install); -5: A shell script downloads and runs code from an unrecognised source at node-version.mjs:27 (install))
The repo's own install script failed after its dependencies installed (the repo's own lifecycle script failed (exit 1): . preinstall: Failed); what ran before it was observed and scored
Entry phase did not complete (failed), so later behaviour was not observed
Key-shaped strings in 154 files, never seen leaving the machine
Reputation
Warning: the owner has little public history.
Reputation warnings are shown on their own and do not lower the code score.
Shown apart from the code and behaviour findings.