backlog-orchestrator

Drive a backlog of GitHub issues to PRs autonomously and sequentially. The loaded session becomes an orchestrator that triages issues into S/M/L tiers, spawns PAW implementer (and optional PAW Review) worker terminals that coordinate over telex instead of GitHub-comment polling, gates each PR through a preference/human-floor merge review, and auto-merges or routes to human review. Use when asked to work through a backlog/queue of issues autonomously, run an autonomous issue-fixing pipeline, or orchestrate PAW sessions across many issues.

Jump to install

Repository: lossyrob/skills
Last source activity: July 9, 2026 at 21:33
Detected SKILL.md language: English
Stars: 4
Forks: 0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

A direct command skips the review prompt. Inspect the source before running it.

npx skills add https://github.com/lossyrob/skills --skill backlog-orchestrator

The command stays on one line. Scroll horizontally to inspect it before copying.

Prefer a local copy? Download the files currently available to SkillsMP.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

File Explorer

8 files

SKILL.md

Source instructions · Read-only preview

Back to search Back to installation

Backlog Orchestrator

Use this skill when the user wants to autonomously work through a backlog of GitHub issues: each issue is taken to a pull request by an autonomous PAW implementer session, optionally reviewed by an autonomous PAW reviewer session, and then either auto-merged or routed to human review by you, the orchestrator. The whole run is sequential — one issue end-to-end at a time — until every selected issue's PR is in a terminal state (merged, pending human review, or blocked).

You are the orchestrator. You do not write feature code. You triage, spawn workers, coordinate them over telex, gate merges, merge, and report.

The model in one paragraph

Worker sessions run in their own terminal windows/tabs (iTerm2/Terminal.app on macOS or Windows Terminal, via launch-copilot-terminal) and in their own git worktrees. On macOS, automatic terminal selection uses Terminal.app; iTerm2 is an explicit run-level choice because its first launch requires one-time macOS Automation consent. The runs the PAW workflow (config , e.g. , loaded as a skill — launched as the agent) → opens a PR. If the tier calls for it, a — launched (), with its prompt authorizing autonomous review submission — runs the PAW Review workflow → submits a real GitHub PR review. The implementer and reviewer hand off the review cycle (not GitHub-comment polling loops): "review ready" → "review posted" → "re-review requested" → "+1". The implementer keeps using the to repair CI failures and merge conflicts (this stays a loop — telex does not replace it). When the PR is merge-ready, the implementer and messages . You run a last-line (a subagent, not a correctness review) that reads that field report, then either squash-merge (and ) or route to . On a human-review route you do stand the worker down: you send so the implementer's sentry keeps the PR mergeable until the merges; the implementer then messages you , and only then do you stand it down. Either way, you advance to the next issue.