- Home
- Skills
- Agents & Orchestration
- Drain Issues — Dependency-Aware Backlog Execution
Drain Issues — Dependency-Aware Backlog Execution
Process a GitHub backlog in dependency-aware waves with isolated worktrees, independent reviews, CI gates and sequential authorized merges.
$30
Works with the AI Agents you already use
Drain Issues — Dependency-Aware Backlog Execution
Example session with this skill installed
Dry-run this synthetic backlog as example-author with two parallel writers: #11 independent auth change; #12 depends on #11; #13 independent docs change; #14 assigned to another author; #15 tracking-only epic; #16 and #17 depend on each other; #18 depends on unavailable #99. Do not assign, edit, push or merge anything.
- Read your context and instructions
- Compiled the drain issues
Synthetic backlog planning demonstration
The packaged plan_waves.py helper was executed locally against this normalized inventory. It returned exit code 0 and this complete JSON result:
{"waves":[[11,13],[12]],"blocked":{"16":"dependency cycle or descendant of a cycle","17":"dependency cycle or descendant of a cycle","18":"unsatisfied excluded/missing dependency: 99"},"skipped":{"14":"assigned to another actor","15":"tracking-only; no implementation work"},"notice":"Conditional plan only: later waves require verified prerequisite merges on base."}
Execution meaning
The first potential wave contains #11 and #13. Issue #12 is conditional on #11's change actually being merged and present on its base; a ready PR is insufficient. With no-merge mode, #12 remains blocked until that prerequisite lands. The cycle and missing prerequisite do not prevent the independent wave. Issue #14 retains its existing assignment, and the tracking epic is not treated as implementation work.
Validation scope
This was a local dry-run with synthetic input. The agent must supply correctly interpreted dependencies and complete inventory; the helper does not infer them from prose or access GitHub. No issue, branch, PR or merge was created in this demonstration.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn a scoped GitHub backlog into an orderly sequence of reviewed PRs. Drain Issues distinguishes real prerequisites from tracking checklists, plans independent work in bounded waves, and keeps blocked issues from stalling unrelated work.
Each issue gets its own branch, worktree, reviewed plan, coder handoff and Implementation Report. The root scheduler performs final reviews and merges sequentially, refreshing the base and invalidating stale evidence after changes. Cycles, unavailable prerequisites, assignment conflicts and missing CI remain explicit in the run record. Resume from verified live state without blindly creating duplicate branches or PRs.
Includes the complete per-issue pipeline and review procedure, scheduler guidance, evidence and inventory formats, and two read-only Python helpers: a dependency-wave planner and a readiness-record checker. Neither helper contacts GitHub or executes project code. Your coding agent runs the workflow through its own authorized tools and isolated sessions. Default concurrency is two issue writers, capped at three. Dry-run and no-merge modes are included.
Created from my public project: https://github.com/lhermoso/claude-workflows . The original MIT-licensed drain-issues command is freely available there. This standalone edition bundles the supporting pipeline, replaces private command paths and fixed provider runners with host capability checks, and adds tested planning and evidence helpers. MIT license and attribution are included.
No separate Issue Pipeline purchase is required. This is a workflow for an active coding-agent task, not a hosted service or persistent autonomous daemon. Model usage and infrastructure are not included.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
This is an agent workflow, not a hosted service or an unattended scheduler. Results depend on the host's agent isolation, tool permissions, repository test suite and GitHub policy. AI review can miss defects. Missing CI or independent reviewer capability blocks completion. Model usage and infrastructure costs are separate from this one-time skill purchase. Local helper tests and simulated workflow decisions do not establish live end-to-end reliability.
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 buying a skill.
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
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 26 days ago
- Accounts
GitHub
- Software
Git
Host GitHub toolsorGitHub CLI
Your project's test runtimes- Optional
Python 3.9+
Your agent must be able to start fresh, isolated sessions. Python runs optional helpers.
Permissions
Allowed Hosts
File Scopes
Reads and writes only the selected repository, task-owned isolated worktrees and private run records. Runs authorized project tests and Git operations; uses GitHub for issue/PR data and authorized assignment, commits, pushes, reports and merges. Fresh agent sessions receive task-relevant evidence under the host's controls. Credentials stay in the existing authorized host/CLI integration; never paste them into prompts or logs. No deployment, branch-protection changes, admin bypass or persistent background service. Dry-run performs no project execution or Git/GitHub mutation.