The reachable postinstall hook writes persistent shell completion files and can repoint the package's own Claude skill link, which is a warning-level lifecycle risk. The flagged telemetry is disclosed, bounded to diagnostic fields, and sent to PostHog; I found no concrete credential theft or unrelated exfiltration.
Static reason
High-risk behavior combination matched malicious policy.; source matched previously finalized malicious package; routed for review
Trigger
The npm postinstall hook runs during installation.
Impact
Shell completion files and the Loki skill link persist in the user's home directory after installation.
Mechanism
The hook invokes the Loki completion command and a helper that can replace an existing ~/.claude/skills/loki-mode symlink.
Evidence
package.jsonbin/loki-postinstall.jsloki-ts/dist/loki.jsautonomy/lib/skill-link-heal.sh~/.local/share/bash-completion/completions/loki~/.config/fish/completions/loki.fish~/.zfunc/_loki~/.loki/completions-version~/.claude/skills/loki-mode