sakhi-shraddha-sst Duplicate Bug Detection Agent
Prevents JIRA clutter by identifying duplicate or similar bugs before ticket creation.
When to use this skill
- QA is writing a new bug report
- User mentions "similar issue" or "might be duplicate"
- Before creating a new JIRA bug ticket
- Reviewing backlog for potential duplicates
Workflow
- Collect the new bug summary and description from QA
- Search existing JIRA bugs using relevant keywords
- Fetch full details of candidate duplicate bugs
- Compare similarity using text comparison logic
- Return recommendation with confidence score
Instructions
Step 1: Extract Bug Information
Gather from user input:
- Summary: Brief title of the bug
- Description: Detailed explanation of the issue
- Component (optional): Affected module/feature
- Labels (optional): Existing tags
Step 2: Search Existing Bugs
Construct JQL query to find potential duplicates:
project = * AND (
summary ~ "keyword1 keyword2" OR
description ~ "keyword1 keyword2" OR
text ~ "error message"
) AND status != Closed ORDER BY updated DESC
Search strategies:
- Extract 3-5 key terms from the new bug summary
- Include error messages or codes
- Search within same component if known
- Limit to last 6 months (adjustable)
Step 3: Compare Similarity
For each candidate bug, compare:
| Field | Weight | Method |
|---|---|---|
| Summary | 40% | Cosine similarity or keyword overlap |
| Description | 40% | Semantic similarity, error pattern match |
| Component | 10% | Exact match bonus |
| Steps to Reproduce | 10% | Sequence similarity |
Similarity scoring:
- > 85%: High probability duplicate
- 70-85%: Possible duplicate (flag for review)
- < 70%: Likely unique
Step 4: Return Recommendation
Response format:
Duplicate Check Results:
━━━━━━━━━━━━━━━━━━━━━━━
🔍 Analyzed: [X] existing bugs
✅ HIGH MATCH (>85%):
• BUG-245: "[Summary]" (92% similar)
→ URL: https://jira.company.com/browse/BUG-245
⚠️ POSSIBLE MATCH (70-85%):
• BUG-189: "[Summary]" (78% similar)
→ URL: https://jira.company.com/browse/BUG-189
✨ Recommendation: [Create New / Review BUG-245 / Add Comment to BUG-189]
JIRA Integration
API Endpoints needed:
# Search issues
GET /rest/api/2/search?jql={query}&fields=summary,description,status,key
# Get issue details
GET /rest/api/2/issue/{issueKey}
Authentication:
- Use API token stored in env:
JIRA_API_TOKEN - Base URL from env:
JIRA_BASE_URL(e.g., https://company.atlassian.net)
Scripts
Use the helper script for similarity calculation:
python .agent/skills/detecting-duplicate-bugs/scripts/similarity_checker.py \
--new-bug "bug_summary.txt" \
--candidates "candidate_bugs.json" \
--threshold 0.7
Example Usage
User: "Login button not working on mobile safari"
→ Search JQL: project = * AND (summary ~ "login button" OR summary ~ "mobile safari")
→ Compare with top 20 results
→ Output: "Similar bug found: BUG-178 (85% match) - Login failure on iOS browsers"
Resources
- scripts/similarity_checker.py - Text similarity engine
- examples/duplicate_check_example.md - Sample workflow
Configuration
Set in environment or .env file:
JIRA_BASE_URL=https://yourcompany.atlassian.net
JIRA_API_TOKEN=your_api_token_here
JIRA_PROJECT_KEYS=PROJ,QA,BUG
SIMILARITY_THRESHOLD=0.75