메뉴
Guided introduction to Transformers for new users — what it is, how to use it, what each agent does.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Auto-checks project context, logs activity via scribe, and manages the activity log and artifact directories. Loaded by Optimus and Megatron on every command.
Silent logger skill. Provides Scribe agent with rules for activity logging, memory writes, and report generation within .claude/transformers/.
Deep git and VCS knowledge — worktrees, branching, rebase, conflict resolution, PR review, bisect, release workflow.
Decision intelligence system — agents learn decision rules from every task. Not a cache. Only stores what changes future behavior.
How to create effective Transformers skills — structure, progressive disclosure, bundled resources, testing.
Pre-completion self-check for all builder agents. Run before marking any task done.
| name | onboarding |
| description | Guided introduction to Transformers for new users — what it is, how to use it, what each agent does. |
| user-invocable | true |
Transformers is multi-agent orchestration for serious software development. You get a team of named specialist agents, coordinated by two orchestrators. Not a chatbot. Not a single assistant. A team.
Run /transformers:init. It creates a project context file and loads your stack and structure. That's it. Start working.
Optimus Prime — leader of the Autobots. Builder. Say optimus, build X or optimus, bugfix Y. He decomposes the work, delegates to specialists, manages approval gates, and delivers. He never writes code directly — he coordinates the team that does.
Megatron — leader of the Decepticons. Tester. Say megatron, test X. He deploys specialist testers from multiple angles simultaneously. He finds what breaks before it reaches production.
Build a feature:
optimus, build [feature description]
Full lifecycle: gather requirements, research, plan, build, review, test. Two human approval gates before any code is written.
Fix a bug:
optimus, bugfix [description]
Gather, investigate, plan, fix, verify.
Quick task — call a specialist directly:
ironhide, add a rate-limited endpoint for /api/export
bumblebee, fix the alignment on the dashboard header
ratchet, write a migration to add soft deletes to orders
Test an area:
megatron, test the checkout flow
megatron, security audit the auth module
Git work:
gitmaster, review this PR
gitmaster, set up worktrees for the auth refactor
Debug a crash:
/transformers:debug [error message or description]
Research before building:
/transformers:research [topic or question]
Transformers gets smarter per project. After each session, agents write decision rules to memory — what patterns work, what to avoid, project-specific workarounds. An index of long-term memory is loaded on every command.
After a few sessions, agents know your project. They'll know which directories to look in, which patterns you use, which gotchas exist. You don't have to re-explain.
Structured, multi-phase development work. Building features with proper planning and approval gates. Fixing bugs systematically. Running multi-angle test campaigns. Code review with a specialist reviewer.
If you're making a quick edit to one file, you don't need orchestration. If you're building something real — new feature, complex fix, comprehensive tests — this is the tool. The orchestrators decompose the work so you don't have to.