Agents

Agents (brewers) are specialized roles defined in markdown files with their own model, toolset and system prompt. Brewcode ships 7 (5 public + 2 internal helpers). The five public agents are specialists only: skills, agents, hooks, mods and shell scripts.

Generic roles (developer, tester, reviewer, architect) are not shipped. Every project generates its own into .claude/agents/ with /brewcode:teams-setup or /brewcode:agents; skills that need a reviewer or implementer pick the project’s agent, falling back to general-purpose. Brewcode skills spawn the relevant agent automatically via the Agent tool.

Summary table

Agent	Model	Write tools	Description (from frontmatter)
skill-creator	inherit	Read, Write, Edit, Bash	Creates and improves Claude Code skills. Triggers: create skill, improve skill, fix activation.
agent-creator	inherit	Read, Write, Edit, Bash	Creates and improves Claude Code agents. Triggers: create agent, improve agent, scaffold agent.
hook-creator	inherit	Read, Write, Edit, Bash	Creates and debugs Claude Code hooks. Triggers: create hook, PreToolUse hook, debug hook.
bash-expert	inherit	Read, Write, Edit, Bash	Creates sh/bash scripts for Mac/Linux. Triggers: create script, bash script, shell script.
mod-creator	inherit	Read, Write, Edit, Bash, Skill	Creates, tests and ships Claude Code mods (hooks-module plugins). Triggers: create mod, hooks module, mod API.
bc-rules-organizer	haiku	Read, Write, Edit, Bash	Internal. Spawned only by /brewcode:rules and /brewcode:convention-setup. No direct/auto use.
setup-upgrader	inherit	Read, Write, Edit, Bash, Agent	Internal. Spawned only by /brewcode:setup-status upgrade. No direct/auto use.

inherit runs on the session’s model unless the spawn sets one, so you control the tier. Only the internal bc-rules-organizer is pinned to haiku (cheap file reorganization). The description string is what Claude matches against when auto-selecting an agent.

brewcode mods: Prompt placeholders · Plan-mode effort · Reminders · all mods

Technical details

▶ Scope guard: shared by all 5 public agents #

Every public brewcode agent sizes the task before starting. One bounded unit = one deliverable, roughly 5 files, roughly 10 steps. Anything larger is refused up front.

Situation	What the agent does
Brief fits one bounded unit	Starts immediately
Brief exceeds ~5 files / ~10 steps, or bundles independent deliverables	Stops before starting, returns a split proposal: 2-N bounded subtasks, each with scope and a suggested owner
Scope grows mid-flight	Stops at the next clean boundary, reports done / remaining / how to split
Brief lacks GOAL, SCOPE, CONTEXT, CONSUMER or acceptance	States the assumption in the report, or asks once; never invents scope

Agents deliver for the CONSUMER, not the literal wording: the result must be usable as-is by whoever picks it up next. A split proposal means the guard worked. Re-send one subtask, or spawn the subtasks as parallel agents in a single message.

▶ Creator contract suite #

agents/tests/suite-creator-contract.mjs is a 120-check regression suite that pins what hook-creator, agent-creator, skill-creator and mod-creator teach about the Claude Code hook, subagent and mod API, so an upstream change fails a test instead of silently teaching a stale one.

bc-rules-organizer is internal: spawned only by /brewcode:rules and /brewcode:convention-setup. Agents text-optimizer, ssh-admin and deploy-admin live in Brewtools.

Related

Update: /brewtools:plugin-update, /reload-plugins, then /brewcode:setup-status upgrade - FAQ