- Home
- Skills
- Agents & Orchestration
- backlog cleanup
backlog cleanup
Groom GitHub Issues by identifying completed work, duplicates, and stale labels using real code evidence.
Free
Works with the AI Agents you already use
backlog cleanup
Example session with this skill installed
Run a backlog cleanup on this repo. Show me what's already done or stale.
- Read your context and instructions
- Compiled the backlog cleanup
I have analyzed the open issues against the current codebase and git history. I found 4 issues that appear to be completed, redundant, or stale.
Backlog Grooming Proposal
| Issue | Action | Reason |
|---|---|---|
| #42: Add Prometheus metrics endpoint | Close | Implemented in src/monitoring/metrics.go (line 12). |
| #55: Fix header padding on mobile | Close | Resolved by commit a1b2c3d which addressed mobile CSS. |
| #39: User login refactor | Merge into #31 | Fully redundant; #31 covers the same auth overhaul. |
| #22: Implement dark mode | Remove Label | in progress since 20 days ago with no commits or PRs found. |
| #61: [Closed] Setup CI | Remove Label | Issue is closed but still carries the in progress label. |
Next steps
- Reply with the issue numbers you want me to process (e.g., "Do all" or "42, 39").
- I will close the completed issues with evidence comments and strip stale labels.
- If you're using Claude Code, you can view the updated state in Sideboard.
Connects securely to your tools. The creator never sees your data.
About this skill
Maintaining a clean issue tracker shouldn't be a manual chore. This skill grooms your GitHub repository by identifying issues that no longer belong in the backlog based on the actual state of your codebase. It cross-references open issues against your files, commits, and activity logs to find finished work that was never closed.
What it does
- Verify implementations by scanning code for file:line evidence, functions, or endpoints that prove an issue is already done.
- Detect redundancies by finding duplicate or overlapping issues across both open and closed states.
- Audit stale progress by identifying "in progress" labels on issues with no commits or PR activity for 14 days.
- Cleanup labels by automatically finding closed issues that still carry active progress tags.
- Batch propose changes via a structured checklist for your approval before any write actions occur.
How it works
- Fetch state from the GitHub CLI, retrieving the latest 500 open issues and any closed issues with active labels.
- Scan the codebase using
git logand file inspection to find evidence of completed work or recent activity. - Draft a proposal listing specific issues to close, merge, or unlabel with a concrete reason for each.
- Execute updates once you approve, using the
ghtool to update issue statuses and labels.
Frameworks & tools
This skill requires the GitHub CLI (gh), git, and an authenticated session. It works with any repository regardless of language or framework.
Why this beats prompting it yourself
Generic prompts can't verify code evidence or calculate label staleness accurately. This skill uses specific git log grep patterns and gh api event timelines to ensure every suggestion is backed by hard data rather than LLM hallucinations.
Use cases
- Post-sprint cleanup to close issues that were accidentally left open after a merge.
- Duplicate hunting to consolidate feedback during high-volume periods.
- Backlog grooming to remove progress labels from tasks that were abandoned or paused.
Known limitations
Requires local git checkout and GitHub CLI authentication. It cannot see private data outside the repository scope or analyze third-party project boards not integrated with GitHub Issues.
One of five backlog skills that work together: suggest, order, next, keeper, cleanup. From the maker of Sideboard, a live GitHub Issues board for Claude Code.
Requires GitHub CLI (gh) and local git repository. Identifies activity based on commit messages and PR links; may miss work not referenced by issue number.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account