dev-complete

Orchestrates the full local development loop from raw request to local commits.

Updated Mar 9, 2026
One-click install
npx skills add https://github.com/GinoCanessa/dotnet-fhir-packages --skill dev-complete-ginocanessa
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dev-complete
Source: https://github.com/GinoCanessa/dotnet-fhir-packages/tree/main/.github/skills/dev-complete
Command: npx skills add https://github.com/GinoCanessa/dotnet-fhir-packages --skill dev-complete-ginocanessa

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running a feature request or bug report through the full development workflow normally requires manually invoking a separate skill at each stage — authoring, approach selection, planning, execution, and review — and pausing to answer questions between stages. This Skill drives that entire chain in one invocation, resolving open questions into recorded assumptions instead of stopping. ## Core Features & Use Cases - Fixed stage chain orchestration: Sequentially dispatches dev-request or dev-report, dev-approach, dev-plan, dev-do, and a dev-review remediation tail, one sub-agent per stage. - Autonomous question resolution: Answers stage prompts on the merits from the source, repository, and AGENTS.md, recording each decision as an ASSUMPTION ledger line in the owning artifact. - Resumable runs: Re-invocation with the same arguments resumes at the first incomplete stage, judged by status markers on disk, and rebuilds the assumption ledger from artifacts. - Use Case: Hand the Skill a slot number, the kind 'request', and a GitHub issue reference; it authors the feature request, generates and judges approaches, writes and executes the plan, runs review remediation, and closes with a report of every assumption made — committing locally only, never pushing or opening pull requests. ## Quick Start Run the dev-complete skill on slot 2 as a request with the content 'gh#42' to carry that issue from raw input to local commits in one invocation.

Frequently Asked Questions about dev-complete

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a feature request end-to-end without invoking each dev skill manually?

Invoke dev-complete with a slot number or slot directory path, the kind 'request', and the raw content or issue reference. It dispatches dev-request, dev-approach, dev-plan, dev-do, and a dev-review remediation tail in order, ending with local commits.

How does dev-complete handle questions a stage would normally ask the user?

It resolves each question on the merits using the source, repository, and AGENTS.md, then records the answer as an ASSUMPTION ledger line in the stage's own artifact. Hand-off and publish offers are always declined, and safety gates like undocumented build commands remain blockers.

Can a dev-complete run be resumed after it hands back?

Yes. Re-invoking with the same arguments resumes at the first incomplete stage, judged by status markers in the artifacts rather than file presence. The assumption ledger is rebuilt from disk before continuing, so the closing report covers assumptions made before and after the hand-back.

Does dev-complete push commits or open pull requests on GitHub?

No. The run commits locally only and never pushes, opens pull requests, or writes to GitHub. Publishing stays user-initiated through the separate dev-issue and dev-pr-open skills.

What do max_subagents and review_iterations control in dev-complete?

max_subagents (default 3, cap 8) is a concurrency ceiling passed unchanged to stages that accept it. review_iterations (default 1, cap 5) sets how many dev-review and remediation cycles run after execution; setting it to 0 ends the run at dev-do.

Why does dev-complete stop instead of answering some questions?

Certain conditions are blockers, not questions: an undocumented build or test command, dev-do's safety yields such as a dirty git index, and a slot holding a source artifact of the other kind. These hand back with a named reason and a resume command.