From microfactory
Claims and implements the next microfactory issue — autofixes, tests, pushes, and updates the issue. Use as a /loop target for continuous development.
How this skill is triggered — by the user, by Claude, or both
Slash command
/microfactory:implement-nextThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
One implementation iteration: claim → implement → test → push → update the issue.
One implementation iteration: claim → implement → test → push → update the issue.
Read .microfactory/config.yaml in the project root. If it does not exist, tell the user to run /microfactory:init-factory and stop.
Load the backend skill matching backend (jira-tasks, github-tasks, or todo-tasks) for all task operations below.
In Progress (best effort) instead of searching.Use a feature branch when feature_branches: true in config or the issue has a needs-branch label; a skip-branch label overrides both and forces direct-to-default. When branching, create feature/<KEY> from the latest default branch (reset it to the default branch if it already exists).
plans/<KEY>.md exists, it is the approved plan — follow it.Feature-branch mode:
[<KEY>] <summary>, body referencing the issue.In Review.Direct mode:
Implemented: <summary> on the issue and transition it to Done.Issue comments and transitions after the code is pushed are best effort: if they fail, warn and finish — the pushed code is the important artifact.
If you cannot complete the implementation (e.g. tests cannot be made to pass), do not transition the issue to Done: push nothing broken, comment on the issue describing what blocked you, and leave it In Progress for a human.
Handle one issue per iteration.
npx claudepluginhub jaksa76/microfactoryRuns one planning iteration of the microfactory: claims an issue, explores the codebase, writes a plan, pushes it, and requests review. Designed for use as a /loop target.
Orchestrates a complete development cycle from GitHub Issue to merged PR: parse issue, explore codebase, plan, execute with agents, create PR, and cleanup. Use via /issue-flow.
Implements a feature from the issue tracker by dispatching specialist subagents for specification, design, code analysis, review, testing, and publishing. Supports decompose, dry-run, yolo, and step modes.