|
1 | 1 | # Parallel Explore Agents for Large Data Processing |
2 | 2 |
|
3 | | -**Status:** Implemented via 3-step routing in `workflow_orchestrator.md`. Read-only breadth tasks are handled by spawning parallel Explore agents or codebase-context-analyzer directly. |
| 3 | +**Status:** Implemented via 3-step routing in `commands/delegate.md`. Read-only breadth tasks are handled by spawning parallel Explore agents or codebase-context-analyzer directly. |
4 | 4 |
|
5 | 5 | ## Background |
6 | 6 |
|
@@ -95,7 +95,7 @@ The distinction: **depth** tasks go to specialized agents, **breadth** tasks go |
95 | 95 |
|
96 | 96 | ## Changes Required |
97 | 97 |
|
98 | | -### 1. `system-prompts/workflow_orchestrator.md` (Planning Instructions section) |
| 98 | +### 1. `commands/delegate.md` (Planning Instructions section) |
99 | 99 |
|
100 | 100 | **Add Explore to Available Specialized Agents table:** |
101 | 101 |
|
@@ -128,7 +128,7 @@ REQUIRED: Large data processing workflows MUST include a final aggregation wave. |
128 | 128 |
|
129 | 129 | Add: "Explore agents use Haiku (cheap, fast). Higher concurrency is acceptable for Explore-only waves." |
130 | 130 |
|
131 | | -### 2. `system-prompts/workflow_orchestrator.md` |
| 131 | +### 2. `commands/delegate.md` |
132 | 132 |
|
133 | 133 | **Add Explore Agent Execution section:** |
134 | 134 |
|
@@ -215,7 +215,7 @@ Example Task invocation for Explore: |
215 | 215 |
|
216 | 216 | ### Implementation (Completed) |
217 | 217 |
|
218 | | -1. **3-step routing in `workflow_orchestrator.md`** |
| 218 | +1. **3-step routing in `commands/delegate.md`** |
219 | 219 | - Step 1: Write detection (skip direct Explore routing if write indicators found) |
220 | 220 | - Step 2: Breadth task detection (single verb + broad scope) |
221 | 221 | - Step 3: Route decision (parallel Explore agents, plan mode, or direct execution) |
|
0 commit comments