-
Notifications
You must be signed in to change notification settings - Fork 1
Closed
Labels
create-expertcreate-expert CLI packagecreate-expert CLI packagerefactorCode improvement without behavior changeCode improvement without behavior change
Description
Description
The integration-manager expert has multiple responsibilities that may violate Best Practice #1 "Do One Thing Well".
Current State
integration-manager does:
- Parallel orchestration - Runs functional-manager and usability-manager simultaneously
- Results collection - Waits for and gathers reports from both managers
- Trade-off analysis - Identifies conflicts between functional and usability requirements
- Integration verification - Verifies ecosystem experts work together
- Holistic assessment - Calculates combined quality scores
instruction = '''
## Workflow
### 1. Parallel Testing
Delegate to both managers simultaneously...
### 2. Collect Results
Wait for both managers...
### 3. Trade-off Analysis
Identify any conflicts...
### 4. Integration Verification
Verify ecosystem experts work together...
### 5. Holistic Assessment
Calculate overall quality score...
'''Analysis
Per Best Practice #1:
Experts that do everything eventually break under their own weight.
Counter-argument from the code:
## Architecture Note
The 4-level delegation depth (create-expert → integration-manager → functional/usability-manager → expert-tester)
is intentional for separation of concerns
The 4-level depth is intentional, but the question is whether integration-manager itself is too broad.
Questions to Consider
-
Is the current design correct?
- integration-manager as "integration" coordinator makes sense
- Trade-off analysis is integration-specific
- Maybe this is the right level of abstraction
-
Could responsibilities be split?
- Separate
quality-scorerfor calculating scores - Separate
trade-off-analyzerfor conflict analysis - Keep integration-manager focused on orchestration only
- Separate
-
What's the actual problem?
- Is the instruction too long?
- Is the expert unreliable in practice?
- Or is this just theoretical concern?
Recommendation
This issue is for discussion. The current design may be acceptable if:
- integration-manager is still focused (all responsibilities relate to "integration")
- The expert performs reliably
- The instruction length is manageable
Mark as "needs triage" for team discussion.
Affected Areas
apps/create-expert/src/lib/create-expert-toml.ts(INTEGRATION_MANAGER_INSTRUCTION)
Acceptance Criteria
- Team decision: keep current design or split
- If splitting: define new expert boundaries
- If keeping: document why this is appropriate exception to Principle Docs: Add related documentation links to runtime README #1
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
create-expertcreate-expert CLI packagecreate-expert CLI packagerefactorCode improvement without behavior changeCode improvement without behavior change