PLUGIN brewcode / Skills
On this page
Skills
Skills (recipes) are slash commands defined in SKILL.md files. Brewcode ships 9, from project analysis to E2E testing. A tenth is generated into every target project by /brewcode:superreview-setup and invoked there as /superreview (see the emitted skill page).
Summary table
| # | Skill | Model | Purpose |
|---|---|---|---|
| 1 | /brewcode:agents | opus | Creates, improves, syncs Claude Code subagents |
| 2 | /brewcode:convention-setup | opus | Coding, testing and architecture conventions with reversible loading |
| 3 | /brewcode:e2e | opus | E2E orchestration: BDD scenarios, Playwright autotests |
| 4 | /brewcode:rules | sonnet | Syncs KNOWLEDGE.jsonl or session learnings to project rules |
| 5 | /brewcode:semble-setup | opus | Installs, audits and maintains the semble_code search MCP |
| 6 | /brewcode:setup-status | sonnet | State of 11 setups and 23 carriers (brewcode 7, brewtools 13, brewdoc 3) with next commands; upgrade brings the stale ones current |
| 7 | /brewcode:skills | opus | Lists, improves, creates, syncs Claude Code skills |
| 8 | /brewcode:superreview-setup | opus | Generates a project-tailored deep-review skill |
| 9 | /brewcode:teams-setup | opus | Creates and manages dynamic teams of domain agents |
| 10 | /superreview (emitted) | opus | Emitted, not shipped: per-project deep review, written into your repo by superreview-setup |
brewcode mods: Prompt placeholders · Plan-mode effort · Reminders · all mods
The -setup naming rule
Read the suffix, know what you get
A -setup skill installs a mechanism (agents, hooks, a rule, a generated skill) that you use afterwards instead of the skill itself. Run it once, then upgrade. Bare-named skills are recurring tools and leave no installed state.
All eleven -setup skills implement the same verbs, in this order: status · install · upgrade · enable · disable · uninstall · purge. Extras come after, never in place (reindex, optimize, resume on semble-setup; [name] on teams-setup). A bare call means status when installed and install when absent. Two groups differ: semble-setup (installation reaches outside the project) and the four brewtools guard setups (manager-setup, agent-deadline-setup, agent-return-setup, agent-router-setup) always default to status; convention-setup defaults to install.
Run order: setup-status, semble-setup, convention-setup, teams-setup, superreview-setup; the full chain is under Skill dependencies. setup-status, superreview-setup, skills, agents and e2e also work standalone. Run teams-setup upgrade later to tune underperforming agents from recorded trace data. text-optimize, text-human, secrets-scan, ssh and deploy live in the Brewtools plugin.
Technical details
Delegation: one contract for skills that spawn agents #
Every skill here except semble-setup (it has no Agent tool) fans work out to subagents under one sizing rule; setup-status upgrade spawns one setup-upgrader per stale setup. A big task handed to one agent means an agent gone for an hour that you cannot observe or correct, and it usually drifts.
One subagent = one bounded unit: one deliverable, roughly 5 files, roughly 10 steps. Larger work splits into independent parallel assignments. Convention setup partitions layer analysis, E2E partitions scenarios, teams setup creates agents one file per spawn.
Every spawn prompt carries six fields; a bare one-line task is never enough:
| Field | Content |
|---|---|
| GOAL | the overall task and why it exists, beyond the file edit |
| ROLE | what this agent owns and must NOT touch |
| SCOPE | exact paths/commands in bounds plus explicit out-of-bounds |
| CONTEXT | what is already done, by whom, what runs in parallel, trimmed to this agent |
| CONSUMER | who or what uses the result next, and the shape it must fit |
| DONE | acceptance criteria plus the exact report shape wanted back |
The same brief lives in E2E references, convention-setup’s rules guide, teams-setup’s roster template and superreview-setup’s emitted template. Main owns spawns and integration; delegates return decisions and results rather than nesting. Agents and skills management default to read-only status, use existing answers, resolve explicit modes anywhere in the prompt and bundle unresolved write decisions into one pre-write question.
Skill dependencies #
setup-status (what is already here; upgrade refreshes the stale ones)
|
semble-setup (search foundation, when applicable)
|
convention-setup (coding, architecture and testing foundation)
|
teams-setup --> project domain agents
|
superreview-setup --> /superreview (emitted into your repo)
rules / skills / agents / e2e remain recurring toolsRelated
Brewcode overview
Plugin skills, agents and hooks.
Brewcode mods
Add-ons: prompt placeholders and plan-mode effort.
View on GitHub
Source code and READMEs.
Update: /brewtools:plugin-update, /reload-plugins, then /brewcode:setup-status upgrade - FAQ