1---2name: ultrawork-53description: Parallel execution engine for high-throughput task completion4---5
6<Purpose>
7Ultrawork is a parallel execution engine that runs multiple agents simultaneously for independent tasks. It is a component, not a standalone persistence mode -- it provides parallelism and smart model routing but not persistence, verification loops, or state management.
8</Purpose>
9
10<Use_When>
11- Multiple independent tasks can run simultaneously
12- User says "ulw", "ultrawork", or wants parallel execution
13- You need to delegate work to multiple agents at once
14- Task benefits from concurrent execution but the user will manage completion themselves
15</Use_When>
16
17<Do_Not_Use_When>
18- Task requires guaranteed completion with verification -- use `ralph` instead (ralph includes ultrawork)
19- Task requires a full autonomous pipeline -- use `autopilot` instead (autopilot includes ralph which includes ultrawork)
20- There is only one sequential task with no parallelism opportunity -- delegate directly to an executor agent
21- User needs session persistence for resume -- use `ralph` which adds persistence on top of ultrawork
22</Do_Not_Use_When>
23
24<Why_This_Exists>
25Sequential task execution wastes time when tasks are independent. Ultrawork enables firing multiple agents simultaneously and routing each to the right model tier, reducing total execution time while controlling token costs. It is designed as a composable component that ralph and autopilot layer on top of.
26</Why_This_Exists>
27
28<Execution_Policy>
29- Fire all independent agent calls simultaneously -- never serialize independent work
30- Always pass the `model` parameter explicitly when delegating
31- Read `docs/shared/agent-tiers.md` before first delegation for agent selection guidance
32- Use `run_in_background: true` for operations over ~30 seconds (installs, builds, tests)
33- Run quick commands (git status, file reads, simple checks) in the foreground
34</Execution_Policy>
35
36<Steps>
371. **Read agent reference**: Load `docs/shared/agent-tiers.md` for tier selection
382. **Classify tasks by independence**: Identify which tasks can run in parallel vs which have dependencies
393. **Route to correct tiers**:
40 - Simple lookups/definitions: LOW tier (Haiku)
41 - Standard implementation: MEDIUM tier (Sonnet)
42 - Complex analysis/refactoring: HIGH tier (Opus)
434. **Fire independent tasks simultaneously**: Launch all parallel-safe tasks at once
445. **Run dependent tasks sequentially**: Wait for prerequisites before launching dependent work
456. **Background long operations**: Builds, installs, and test suites use `run_in_background: true`
467. **Verify when all tasks complete** (lightweight):
47 - Build/typecheck passes
48 - Affected tests pass
49 - No new errors introduced
50</Steps>
51
52<Tool_Usage>
53- Use `Task(subagent_type="oh-my-claudecode:executor", model="haiku", ...)` for simple changes
54- Use `Task(subagent_type="oh-my-claudecode:executor", model="sonnet", ...)` for standard work
55- Use `Task(subagent_type="oh-my-claudecode:executor", model="opus", ...)` for complex work
56- Use `run_in_background: true` for package installs, builds, and test suites
57- Use foreground execution for quick status checks and file operations
58</Tool_Usage>
59
60<Examples>
61<Good>
62Three independent tasks fired simultaneously:
63```
64Task(subagent_type="oh-my-claudecode:executor", model="haiku", prompt="Add missing type export for Config interface")
65Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="Implement the /api/users endpoint with validation")
66Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="Add integration tests for the auth middleware")
67```
68Why good: Independent tasks at appropriate tiers, all fired at once.
69</Good>
70
71<Good>
72Correct use of background execution:
73```
74Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="npm install && npm run build", run_in_background=true)
75Task(subagent_type="oh-my-claudecode:executor", model="haiku", prompt="Update the README with new API endpoints")
76```
77Why good: Long build runs in background while short task runs in foreground.
78</Good>
79
80<Bad>
81Sequential execution of independent work:
82```
83result1 = Task(executor, "Add type export") # wait...
84result2 = Task(executor, "Implement endpoint") # wait...
85result3 = Task(executor, "Add tests") # wait...
86```
87Why bad: These tasks are independent. Running them sequentially wastes time.
88</Bad>
89
90<Bad>
91Wrong tier selection:
92```
93Task(subagent_type="oh-my-claudecode:executor", model="opus", prompt="Add a missing semicolon")
94```
95Why bad: Opus is expensive overkill for a trivial fix. Use executor with Haiku instead.
96</Bad>
97</Examples>
98
99<Escalation_And_Stop_Conditions>
100- When ultrawork is invoked directly (not via ralph), apply lightweight verification only -- build passes, tests pass, no new errors
101- For full persistence and comprehensive architect verification, recommend switching to `ralph` mode
102- If a task fails repeatedly across retries, report the issue rather than retrying indefinitely
103- Escalate to the user when tasks have unclear dependencies or conflicting requirements
104</Escalation_And_Stop_Conditions>
105
106<Final_Checklist>
107- [ ] All parallel tasks completed
108- [ ] Build/typecheck passes
109- [ ] Affected tests pass
110- [ ] No new errors introduced
111</Final_Checklist>
112
113<Advanced>
114## Relationship to Other Modes
115
116```
117ralph (persistence wrapper)
118 \-- includes: ultrawork (this skill)
119 \-- provides: parallel execution only
120
121autopilot (autonomous execution)
122 \-- includes: ralph
123 \-- includes: ultrawork (this skill)
124```
125
126Ultrawork is the parallelism layer. Ralph adds persistence and verification. Autopilot adds the full lifecycle pipeline.
127</Advanced>
majiayu000