Source inspection found an explicit AI-agent scaffolder and an unattended same-package global updater, but no malicious lifecycle chain or data theft. This is a real but bounded user-invoked capability risk, so warn rather than block.
Static reason
No blocking static signals were detected.
Trigger
A user runs a non-help Hedgehog command; AI-agent files are written only through init.
Impact
A command can alter the user's global npm installation and project AI-agent control files; no secret theft, remote payload execution, or unconsented lifecycle execution was confirmed.
Mechanism
Same-vendor global CLI update and explicit project agent scaffolding.
Evidence
package.jsonbin/cli.mjssrc/hosts/index.mjssrc/registry/fetch.mjssrc/hosts/version.mjssrc/db/graph-server.mjs.claude/agents.claude/skills.cursor/agents.cursor/skills.cursor/rules/hedgehog.mdc.gemini/agents.gemini/skillsAGENTS.md