The reachable postinstall hook can update Loki's own Claude Code skill link, which falls under the agent-extension lifecycle risk rule. The behavior is narrowly limited to an existing Loki symlink and a durable install, so this warrants a warning rather than a publish block.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
Runs automatically during npm postinstall when an eligible existing skill symlink points to another Loki installation.
Impact
Claude Code may load the newly installed Loki skill after installation. This is an agent-extension lifecycle risk, but the inspected code does not alter broader agent settings or instructions.
Mechanism
The hook atomically replaces ~/.claude/skills/loki-mode with a symlink to the current durable Loki installation.
Evidence
package.jsonbin/loki-postinstall.jsautonomy/lib/skill-link-heal.sh~/.claude/skills/loki-mode