A county developer wants Claude Code to run city lint and security-scan scripts on demand. What Claude Code feature fits?

Repeatable city lint and security-scan workflows belong in reusable Claude Code commands rather than ad hoc improvisation. Creating custom slash commands that run the city lint and security-scan scripts works because developers invoke the same reviewed command definitions on demand, reducing mistaken one-liners and keeping CI helpers consistent with official scripts. County engineers get speed without inventing shell each time. Configuring a lifecycle hook that runs lint automatically after file edits fails because it runs automatically on edits instead of on demand, and it does not create a reusable command for both the lint and security-scan scripts. Adding the lint and security-scan scripts to CLAUDE.md for project context fails because CLAUDE.md only supplies background context to Claude and does not create an executable on-demand command that runs the scripts. Defining a specialized subagent to audit city code for lint and security findings fails because a subagent is an autonomous assistant rather than a reusable on-demand command bound to the city's exact lint and security-scan entry points. Exam caveat: slash commands should wrap approved scripts from the repo or CI system—custom commands that embed divergent logic recreate shadow pipelines. Operational check: define Claude Code slash commands that call the city’s canonical lint and security-scan entry points, document them in the developer guide, and confirm they match the same scripts CI runs on pull requests.