The default postinstall behavior modifies a project’s Claude skill directory without a separate setup command, which is a warning-level agent extension lifecycle risk. The installed content is the package’s own Ax guidance, and the reviewed path showed no broader agent-setting mutation or install-time data transmission.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Package installation as a direct dependency, unless automatic setup is disabled or a skip condition applies.
Impact
The project gains a package-provided agent skill that may be loaded by Claude; this is a limited lifecycle risk, with no broader agent-setting change identified.
Mechanism
The postinstall script reads the project configuration and writes the bundled Ax skill under .claude/skills.
Evidence
package.jsonscripts/postinstall.mjsscripts/skill-install.mjsskills/ax-llm.md.claude/skills/ax-llm/SKILL.md.claude/skills/ax-llm/references/