Executes a scoped Linear backlog as maximal safe parallel batches via orchestrator child threads.
npx skills add https://github.com/udecode/dotai --skill linear-backlog
Please help me install this Agent Skill. Skill: linear-backlog Source: https://github.com/udecode/dotai/tree/main/skills/linear-backlog Command: npx skills add https://github.com/udecode/dotai --skill linear-backlog
Manually working through a Linear backlog one issue at a time is slow, and naive parallelization causes merge conflicts, duplicated work, and lost state. This Skill runs a frozen Linear queue autonomously as serial batches with maximal safe parallel execution inside each batch, without asking the user to approve every next step.
Ask the agent to use $linear-backlog to run a specific Linear project, cycle, or label as maximal safe parallel batches until no eligible work remains.
High-intent search queries and answers about installing and using this skill.
Use $linear-backlog run with a scope such as a Linear project, cycle, label, saved view, or explicit issue list. The skill freezes the queue, plans maximal safe parallel batches, and dispatches each issue to a durable orchestrator child thread until no eligible work remains.
It builds a conflict graph from dependency-ready issues and only creates hard-conflict edges for real risks like shared migrations, exclusive runtimes, or incompatible API contract changes. Minor file overlap or expected small merge conflicts become conflict groups, not reasons to serialize.
It requires Linear issue read and write tools, the orchestrator skill with durable Codex child-thread tools, autogoal and its goal tools, the destination repo's task skill and AGENTS instructions, plus Git and the repo's normal PR and merge tooling.
Yes. On restart it reads the parent goal and ledger, re-reads the frozen Linear issue ids, inspects durable children, worktrees, branches, and PRs, then reconstructs the active batch before continuing. It never duplicates children, branches, PRs, or goals.
An issue is queue-terminal only when merged, verified, and moved to the team's completed state, when canceled by an authorized source, or when blocked with evidence, an owner, and a concrete next action. Opening a PR or passing tests alone does not count.
Blockers are classified as dependency-local, lane-local, external-owner, scope-authority, or repo-wide. Independent lanes keep running, evidence and next actions are recorded, and the run only stops for scope-authority or repo-wide blockers that affect the whole queue.
LazyIsEfficientAudit Linear/Jira tickets and author ADRs and DADs with lifecycle rules.
Umbral-BotTriage Linear issues into recommendations for project, priority, state, duplicates, owner, and context.
semanticpixelConverts PLAN-*.md files into a GitHub Issues parent and child sub-issue tree.
nthpaulCoordinate dependency-aware parallel task execution with per-lane workers and a durable watcher.
