A self-driving feature lifecycle manager that designs, builds, and merges code autonomously.

Install

mkdir -p .claude/skills/feature-loop && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/19128" && unzip -o skill.zip -d .claude/skills/feature-loop && rm skill.zip

Installs to .claude/skills/feature-loop

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

The autonomous Feature Loop driver. A state machine that owns the whole feature lifecycle — generate 3 candidates from GOAL.md + a recommendation, then (after one human pick) design, slice, build in parallel, review, and auto-merge. Runtime-agnostic: uses Orca workers when Orca is running, else Claude Code sub-agents + git worktrees. Runs one stage per invocation. Reuses /grill-with-docs, /to-spec, /to-tickets, /implement, /code-review. See ADR 0022.
454 chars✓ has a “when” triggerlonger than Claude Code's old 250-char listing cap (fine on current versions)
Advanced

Key capabilities

  • Generate three feature candidates from GOAL.md with a recommendation
  • Design the chosen feature, slice it into tickets, and build in parallel
  • Review the implemented feature and auto-merge it
  • Detect runtime mode (Orca or Local) automatically
  • Run isolated workers for design and parallel build stages
  • Acquire and release a run-lock to prevent overlapping executions

How it works

The skill operates as a state machine, advancing through feature creation stages like candidate generation, design, parallel build, review, and auto-merge. It uses either Orca workers or Claude Code sub-agents for isolated tasks and manages a run-lock.

Inputs & outputs

You give it
a GOAL.md file
You get back
a completed and merged feature

When to use feature-loop

  • Automating feature development
  • Managing feature lifecycle
  • Designing new functionality

About feature-loop

Operates a state machine that handles end-to-end feature creation. It iterates through design, parallel build, and review, using local agents to manage tasks efficiently.

>-

When not to use it

  • When a human operator is available to answer questions mid-run
  • When an ungroundable choice touches a schema/data migration, destructive op, or public contract
  • When the run-lock is already held by another process

Limitations

  • Runs one stage per invocation
  • Escalates to human intervention for irreversible choices
  • Requires a run-lock to prevent concurrent executions

How it compares

This workflow automates the entire feature lifecycle, including parallel builds and auto-merging, and adapts to available runtime environments (Orca or local sub-agents), which is more autonomous than manual feature development processes.

Compared to similar skills

feature-loop side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
feature-loop (this skill)02moReviewAdvanced
cursor-local-dev-loop31moReviewBeginner
codex-worker17moReviewAdvanced
agent-challenges17moNo flagsIntermediate

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry