booster-commandlisted

# Booster Command Runner This skill is the Codex compatibility layer for Claude Booster command specs. Use it when the user invokes a Booster command by name, for example: - `start` - `handover` - `consilium <topic>` - `audit <topic>` - `architecture [--update]` - `go <artifact contract>` - `debt <mode>` - `$consilium <topic>` or `/prompts:consilium <topic>` ## Source Of Truth Load the original command spec before executing. Resolve `<command>` from the alias skill name, prompt name, or the first argument after `$booster-command`. Search in this order: 1. `~/.claude/commands/<command>.md` 2. `references/commands/<command>.md` relative to this installed skill 3. `<repo-root>/templates/commands/<command>.md` 4. `<repo-root>/.claude/commands/<command>.md` If the command spec is missing, say which paths were checked and stop. ## Codex Adapters Execute the command behavior, not the literal Claude Code tool names. - `Read` / `Grep` / `Glob` / `Bash` mean Codex file reads, `rg`, and shell tools. - `EnterPlanMode` / `ExitPlanMode` mean use Codex planning behavior or `update_plan`. - `TaskCreate` / `TaskUpdate` mean use `update_plan` and concise progress updates. - Claude `Agent` / `Explore` / `general-purpose` means Codex subagents when the command explicitly asks for independent agents. The command invocation itself is explicit permission to use those subagents. Evidence is the spawned agent ids or names plus their final messages. If no Codex subagent tool is availa