core agents alignment

This commit is contained in:
Brian Madison
2025-07-12 20:16:05 -05:00
parent f32a5fe08a
commit 24a35ff2c4
16 changed files with 82 additions and 814 deletions

View File

@@ -22,20 +22,7 @@ Create detailed, implementation-ready stories for brownfield projects where trad
## Task Execution Instructions
### 0. Check Workflow Plan and Documentation Context
[[LLM: Check for workflow plan first, then available documentation]]
#### 0.1 Check Workflow Plan
- Load core-config.yaml and check if `workflow.trackProgress: true`
- If yes, check for active plan at `workflow.planFile`
- If plan exists:
- Verify story creation aligns with current plan step
- If out of sequence, warn user (enforce based on config)
- Note which step this story creation corresponds to
#### 0.2 Determine Documentation Context
### 0. Documentation Context
Check for available documentation in this order:
@@ -68,7 +55,7 @@ Based on available documentation:
#### 1.2 Gather Essential Context
[[LLM: For brownfield stories, you MUST gather enough context for safe implementation. Be prepared to ask the user for missing information.]]
CRITICAL: For brownfield stories, you MUST gather enough context for safe implementation. Be prepared to ask the user for missing information.
**Required Information Checklist:**
@@ -78,21 +65,7 @@ Based on available documentation:
- [ ] What technical constraints exist?
- [ ] Are there any "gotchas" or workarounds to know about?
If any required information is missing, ask the user:
```
I need additional context for this brownfield story:
Missing Information:
- [List specific missing items]
Please provide:
1. [Specific question about integration]
2. [Specific question about patterns]
3. [Specific question about constraints]
Or point me to documentation that contains this information.
```
If any required information is missing, list the missing information and ask the user to provide it.
### 2. Extract Technical Context from Available Sources
@@ -117,8 +90,6 @@ If using brownfield PRD:
#### 2.3 From User Documentation
[[LLM: When working with non-BMad documentation, actively extract and organize the information into categories the Dev agent will need]]
Ask the user to help identify:
- Relevant technical specifications
@@ -138,12 +109,13 @@ Start with the story template, filling in what's known:
## Status: Draft
## Story
As a {{user_type}},
I want {{enhancement_capability}},
so that {{value_delivered}}.
## Context Source
[[LLM: Document where story requirements came from]]
- Source Document: {{document name/type}}
- Enhancement Type: {{single feature/bug fix/integration/etc}}
- Existing System Impact: {{brief assessment}}
@@ -151,7 +123,7 @@ so that {{value_delivered}}.
#### 3.2 Develop Acceptance Criteria
[[LLM: For brownfield, ALWAYS include criteria about maintaining existing functionality]]
Critical: For brownfield, ALWAYS include criteria about maintaining existing functionality
Standard structure:
@@ -163,7 +135,7 @@ Standard structure:
#### 3.3 Gather Technical Guidance
[[LLM: This is where you'll need to be interactive with the user if information is missing]]
Critical: This is where you'll need to be interactive with the user if information is missing
Create Dev Technical Guidance section with available information:
@@ -180,22 +152,8 @@ Create Dev Technical Guidance section with available information:
[From documentation or user input]
### Missing Information
[[LLM: List anything you couldn't find that dev will need]]
- [ ] {{missing item 1}}
- [ ] {{missing item 2}}
```
If critical information is missing, pause and ask:
```
To complete the technical guidance for this story, I need:
1. **{{Missing Category}}**:
- Specifically: {{what you need to know}}
- Why needed: {{how it impacts implementation}}
Can you provide this information or point me to relevant code/docs?
```
Critical: List anything you couldn't find that dev will need and ask for the missing information
### 4. Task Generation with Safety Checks
@@ -236,7 +194,7 @@ Example task structure for brownfield:
### 5. Risk Assessment and Mitigation
[[LLM: CRITICAL for brownfield - always include risk assessment]]
CRITICAL: for brownfield - always include risk assessment
Add section for brownfield-specific risks:
@@ -324,15 +282,6 @@ Next Steps:
4. Dev agent can then implement with safety checks
```
### 9. Update Workflow Plan (if applicable)
[[LLM: After successful story creation]]
- If workflow plan tracking is enabled and story was created successfully:
- Call update-workflow-plan task to mark step complete
- Parameters: task: create-brownfield-story, step_id: {from plan check}, status: complete
- If plan shows next recommended step, include it in handoff message
## Success Criteria
The brownfield story creation is successful when:
@@ -348,7 +297,7 @@ The brownfield story creation is successful when:
## Important Notes
- This task is specifically for brownfield projects with non-standard documentation
- Always prioritize existing system safety over new features
- Always prioritize existing system stability over new features
- When in doubt, add exploration and verification tasks
- It's better to ask the user for clarification than make assumptions
- Each story should be self-contained for the dev agent