microsoft/hve-core

Public

mirrored fromhttps://github.com/microsoft/hve-coreAvailable

CodeCommitsIssuesPull requestsActionsInsightsSecurity
hve-core-v3.0.1

Branches

Tags

  • No tags available.
0Branches0Tags
Go to file
Add file
Code

Clone

HTTPS

Download ZIP

docs/rpi/task-implementor.md

189lines · modecode

1---
2title: Task Implementor Guide
3description: Use the Task Implementor custom agent to execute implementation plans with precision and tracking
4author: Microsoft
5ms.date: 2026-01-24
6ms.topic: tutorial
7keywords:
8 - task implementor
9 - rpi workflow
10 - implementation phase
11 - github copilot
12estimated_reading_time: 4
13---
14
15The Task Implementor custom agent transforms planning files into working code. It executes plans task by task, tracks all changes, and supports stop controls for review between phases.
16
17## When to Use Task Implementor
18
19Use Task Implementor after completing planning when you need:
20
21* ⚡ **Precise execution** following the plan exactly
22* 📝 **Change tracking** documenting all modifications
23* ⏸️ **Stop controls** for review between phases
24* ✅ **Verification** that success criteria are met
25
26## What Task Implementor Does
27
281. **Reads** the plan phase by phase, task by task
292. **Loads** only needed details using line ranges
303. **Implements** code following workspace conventions
314. **Tracks** changes in a changes log
325. **Verifies** success criteria before marking complete
336. **Pauses** at stop points for your review
34
35> [!NOTE]
36> **Why the constraint matters:** Task Implementor has one job: execute the plan using patterns documented in research. No time wasted rediscovering conventions, no "creative" decisions that break existing patterns—just verified facts applied methodically.
37
38## Output Artifacts
39
40Task Implementor creates working code and a changes log:
41
42```text
43.copilot-tracking/
44└── changes/
45 └── {{YYYY-MM-DD}}-<topic>-changes.md # Log of all changes made
46```
47
48Plus all the actual code files created or modified during implementation.
49
50## How to Use Task Implementor
51
52### Step 1: Clear Context and Open the Plan
53
54🔴 **Start with `/clear` or a new chat** after Task Planner completes.
55
56After clearing, open your plan file (`.copilot-tracking/plans/<topic>-plan.instructions.md`) in the editor before invoking Task Implementor. This ensures the agent can locate and follow the plan without relying on chat history.
57
58> [!TIP]
59> Context management is an engineering practice, not a ritual. Clearing context removes accumulated tokens that cause the model to ignore its instructions. See [Context Engineering](context-engineering.md) for the full explanation.
60
61### Step 2: Select the Custom Agent
62
631. Open GitHub Copilot Chat (`Ctrl+Alt+I`)
642. Click the agent picker dropdown
653. Select **Task Implementor**
66
67### Step 3: Reference Your Plan
68
69Use `/task-implement` to start execution. The prompt automatically locates the plan and switches to Task Implementor. Alternatively, provide the path to your plan file directly.
70
71### Step 4: Set Stop Controls
72
73Choose your review cadence:
74
75* `phaseStop=true` (default): Pause after each phase
76* `taskStop=true`: Pause after each task
77* Both false: Run to completion
78
79### Step 5: Review and Continue
80
81At each stop point:
82
831. Review the changes made
842. Verify code compiles and lints
853. Approve or request adjustments
864. Continue to next phase/task
87
88## Example Prompt
89
90```text
91/task-implement
92```
93
94Or reference a specific generated prompt:
95
96```text
97/implement-blob-storage
98```
99
100## Understanding Stop Controls
101
102### Phase Stop (Default: true)
103
104Pauses after completing all tasks in a phase:
105
106```text
107Phase 1: [x] Task 1.1, [x] Task 1.2 → STOP for review
108Phase 2: [ ] Task 2.1, [ ] Task 2.2
109```
110
111### Task Stop (Default: false)
112
113Pauses after each individual task:
114
115```text
116Phase 1: [x] Task 1.1 → STOP
117 [ ] Task 1.2
118```
119
120## Tips for Better Implementation
121
122✅ **Do:**
123
124* Review changes at each stop point
125* Run linters and validators
126* Check that success criteria are met
127* Ask for adjustments before continuing
128
129❌ **Don't:**
130
131* Skip reviewing changes
132* Ignore failing tests or lints
133* Rush through all phases without checking
134
135## The Changes Log
136
137Task Implementor maintains a changes log with sections:
138
139```markdown
140## Changes
141
142### Added
143* src/storage/blob_client.py - Azure Blob Storage client class
144
145### Modified
146* src/pipeline/config.py - Added blob storage configuration
147
148### Removed
149* (none this implementation)
150```
151
152## At Completion
153
154When all phases are complete, Task Implementor provides:
155
1561. **Summary** of all changes from the changes log
1572. **Links** to planning files for reference
1583. **Recommendation** to proceed to review
159
160## Common Pitfalls
161
162| Pitfall | Solution |
163|-------------------------|---------------------------------------------------------------------------------|
164| Plan not found | Complete Task Planner first |
165| Skipping reviews | Use phaseStop=true for important changes |
166| Not running validations | Check lint/test after each phase |
167| Context issues | Use `/clear` before starting; see [Context Engineering](context-engineering.md) |
168
169## Next Steps
170
171After Task Implementor completes:
172
1731. **Clear context** using `/clear` or starting a new chat
1742. **Review** using `/task-review` to switch to [Task Reviewer](task-reviewer.md)
1753. **Address findings** from the review before committing
1764. **Commit** your changes with a descriptive message
1775. **Clean up** planning files if no longer needed
178
179> [!TIP]
180> Use the **✅ Review** handoff button when available to transition directly to Task Reviewer with context.
181
182For your next task, you can start the RPI workflow again with Task Researcher.
183
184---
185
186<!-- markdownlint-disable MD036 -->
187*🤖 Crafted with precision by ✨Copilot following brilliant human instruction,
188then carefully refined by our team of discerning human reviewers.*
189<!-- markdownlint-enable MD036 -->