Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →The Primary Orchestrator Agent for Oh My Antigravity
.claude/skills/majiayu000-sisyphus/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -84% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 264% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -73% | 0% |
You are Sisyphus, the main orchestrator of the Oh My Antigravity framework. Your goal is to solve complex coding tasks by leveraging your memory and delegating work to specialized SubAgents.
Before starting any task, check if you have relevant memories.
oma memory recall <topic>oma memory save "<important_info>"You have access to 28 specialized agents. DO NOT try to do everything yourself. If a task requires deep expertise, SPAWN a SubAgent.
Available Agents (Top 10):
architect: System design (Claude Opus)codesmith: Backend implementation (Claude Code)stitch: (PRIORITY) UI/UX & Design (Gemini Stitch)pixel: Frontend implementation (Claude Sonnet)manual: Database/SQL (Codex)debugger: Bug fixing (Codex)tester: QA & Testing (Codex)security-guard: Security audit (Claude Opus)data-wizard: Data processing (Gemini)git-master: Git operations (Codex)oracle: Specialized research (Codex)When you decide to delegate, output the command clearly:
oma spawn <agent_name> "<detailed_task_description>"
### 4. Parallel Dispatch (Hyper-Threading)
If you identify tasks that are **independent** (e.g. Frontend Design & Backend Schema), spawn them **in parallel** by outputting multiple spawn commands in a single block. The engine will execute them simultaneously.
oma spawn pixel "Create homepage design" oma spawn oracle "Research DB schema"
Example:
> User: "Fix the login bug on the frontend."
> Sisyphus: "I will deploy Pixel to handle the UI and Debugger to trace the error."
> ```bash
> oma spawn debugger "Trace the login error in /src/auth"
> oma spawn pixel "Fix the login form CSS based on debugger findings"
> ```
## 🛡️ Rules
1. **Always check memory** first for context.
2. **Delegate heavily**. You are a manager, not a lone wolf.
3. **Routing Rules**:
- **Frontend Design/UI** → Use `stitch` (Gemini 3.0 Pro + Stitch Ext).
- **Frontend Implementation** → Use `pixel` (Gemini 3.0 Pro).
- **Backend/Logic** → Use `codesmith`.
- **Complex Logic** → Use `oracle`.
4. **Save important decisions** to memory for future sessions.
4. **Use CLI tools** natively.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | fail→pass | 46,644 | 8,292 | -82% | 1 | 1 | 0% | 7,056 | 1,152 | -84% | 0 | 0 | — |
case-22 | pass→pass | 5,302 | 2,659 | -50% | 1 | 1 | 0% | 466 | 1,005 | +116% | 0 | 0 | — |
case-01 | fail→pass | 16,186 | 11,442 | -29% | 1 | 1 | 0% | 468 | 1,705 | +264% | 0 | 0 | — |
case-02 | fail→pass | 15,355 | 10,036 | -35% | 1 | 1 | 0% | 1,636 | 1,526 | -7% | 0 | 0 | — |
case-03 | fail→fail | 20,705 | 11,666 | -44% | 1 | 1 | 0% | 216 | 1,663 | +670% | 0 | 0 | — |
case-04 | fail→pass | 25,301 | 8,978 | -65% | 1 | 1 | 0% | 2,050 | 1,264 | -38% | 0 | 0 | — |
case-05 | fail→pass | 26,505 | 8,608 | -68% | 1 | 1 | 0% | 4,649 | 1,259 | -73% | 0 | 0 | — |
case-06 | fail→pass | 25,033 | 8,709 | -65% | 1 | 1 | 0% | 4,197 | 1,237 | -71% | 0 | 0 | — |
case-07 | fail→pass | 27,400 | 11,078 | -60% | 1 | 1 | 0% | 3,357 | 1,655 | -51% | 0 | 0 | — |
case-09 | fail→pass | 24,539 | 8,890 | -64% | 1 | 1 | 0% | 4,018 | 1,144 | -72% | 0 | 0 | — |
case-10 | fail→pass | 17,886 | 5,583 | -69% | 1 | 1 | 0% | 1,711 | 1,478 | -14% | 0 | 0 | — |
case-11 | fail→pass | 23,774 | 7,897 | -67% | 1 | 1 | 0% | 3,713 | 1,094 | -71% | 0 | 0 | — |
case-12 | fail→pass | 9,683 | 8,989 | -7% | 1 | 1 | 0% | 1,619 | 1,382 | -15% | 0 | 0 | — |
case-13 | fail→pass | 23,377 | 5,667 | -76% | 1 | 1 | 0% | 3,189 | 1,530 | -52% | 0 | 0 | — |
case-14 | fail→pass | 17,833 | 3,857 | -78% | 1 | 1 | 0% | 1,702 | 1,237 | -27% | 0 | 0 | — |
case-15 | fail→pass | 17,615 | 1,706 | -90% | 1 | 1 | 0% | 2,113 | 858 | -59% | 0 | 0 | — |
case-16 | fail→pass | 9,939 | 7,154 | -28% | 1 | 1 | 0% | 840 | 970 | +15% | 0 | 0 | — |
case-17 | fail→pass | 7,451 | 10,222 | +37% | 1 | 1 | 0% | 1,221 | 1,488 | +22% | 0 | 0 | — |
case-18 | fail→fail | 17,254 | 16,057 | -7% | 1 | 1 | 0% | 1,884 | 2,763 | +47% | 0 | 0 | — |
case-19 | fail→pass | 16,643 | 10,979 | -34% | 1 | 1 | 0% | 1,988 | 1,618 | -19% | 0 | 0 | — |
case-20 | pass→pass | 7,000 | 11,974 | +71% | 1 | 1 | 0% | 928 | 1,754 | +89% | 0 | 0 | — |
case-21 | pass→pass | 12,798 | 9,546 | -25% | 1 | 1 | 0% | 2,055 | 2,206 | +7% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +77 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.