Skip to content

[plan] Review and fix Issue Arborist workflow effectiveness #10520

@github-actions

Description

@github-actions

Objective

Investigate why the Issue Arborist workflow produced no outputs (0 parent issues, 0 links created) and fix or tune the workflow to produce meaningful results.

Context

The Issue Arborist Daily Report (#10466) shows:

  • Issues Analyzed: 10 (only open issues)
  • Parent Issues Created: 0 (reported as "None")
  • Links Created: 0 (reported as "None")

This suggests the workflow either:

  1. Found no issues meeting criteria for parent creation
  2. Has a bug preventing output generation
  3. Needs dataset expansion or tuning

Root Cause (Hypotheses)

  1. Small dataset: Only 10 open issues analyzed may be too small
  2. Overly strict criteria: Thresholds for creating parent issues may be too high
  3. Logic bug: Workflow may have an error preventing output
  4. Configuration issue: Input parameters may need adjustment

Approach

1. Review Workflow Implementation

Examine the Issue Arborist workflow:

# Find the workflow file
find .github/workflows -name "*issue*arborist*" -o -name "*arborist*"

Check:

  • Input parameters (how many issues to analyze)
  • Criteria for creating parent issues
  • Logic for detecting related issues
  • Output generation code

2. Run Diagnostic Test

Test with different parameters:

# If workflow supports manual trigger with inputs
gh workflow run issue-arborist.lock.yml -f issue_count=50
# Or test locally with workflow file

Try:

  • Increasing issue count (e.g., 50 or 100 instead of 10)
  • Lowering thresholds for parent creation
  • Testing with specific known issue sets

3. Analyze Issue Dataset

Check if the 10 issues analyzed are suitable:

# Get recent open issues
gh issue list --limit 20 --json number,title,labels,createdAt

Questions:

  • Are they related enough to warrant grouping?
  • Do they have common labels/themes?
  • Are they all new or diverse in age?

4. Fix or Tune

Based on findings:

Option A: Expand Dataset

  • Increase default issue count to 50 or 100
  • Include recently closed issues (not just open)
  • Expand time window for analysis

Option B: Adjust Criteria

  • Lower threshold for parent issue creation
  • Adjust similarity/clustering parameters
  • Make criteria more inclusive

Option C: Fix Bugs

  • Fix any logic errors discovered
  • Add error handling for edge cases
  • Improve logging for debugging

Option D: Enhance Logic

  • Improve issue relationship detection
  • Add more sophisticated grouping algorithms
  • Consider label-based or keyword-based clustering

5. Validate Fix

After changes:

  • Trigger workflow manually or wait for next scheduled run
  • Verify report shows non-zero outputs
  • Confirm parent issues and links are meaningful and useful

Files to Modify

  • .github/workflows/issue-arborist.md (workflow definition)
  • Potentially related scripts or actions used by the workflow
  • Run make recompile after modifying .md files

Acceptance Criteria

  • Root cause identified and documented
  • Fix or tuning implemented
  • Workflow produces meaningful outputs (≥1 parent issue or links)
  • Report shows valid analysis results
  • Documentation updated explaining expected behavior

Testing

# Manual test
gh workflow run issue-arborist.lock.yml
# Wait for completion, then check discussion

# Or test locally if workflow supports it
./gh-aw run issue-arborist.md

Success Metrics

AI generated by Plan Command for discussion #10496

Metadata

Metadata

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions