refine

claude-automata

English | 한국어

Plugins that amplify Claude Code's autonomy.

Getting Started

uv is required. If you don't have it, install it first. Runs on POSIX (macOS / Linux / WSL).

Add this repository to the marketplace:

claude plugin marketplace add clomia/claude-automata

Ploop - Advisor Loop

Install: claude plugin install ploop@claude-automata
Update: claude plugin update ploop@claude-automata

ploop is an advisor loop built for long-running work that spans days.

• An independent advisor manages progress on the user's behalf.
  • The advisor finds what the main agent missed.
• The main agent is an orchestrator — it delegates work to agents and stays in command.
• It never loses context across repeated auto-compactions.
  • When a compaction occurs, the anchor is re-injected.
  • The advisor keeps the full context in files.
• It creates no separate sessions and uses only the official subagent path — safe on subscription plans.

The anchor is the file the loop is anchored to. It comes in two kinds.

• Mission (a goal) — receive requirements, process them, and finish once the goal is fully met. Write one with /ploop:define-mission.
• Purpose (a direction) — create requirements as you go and keep advancing, with no fixed end. Write one with /ploop:define-purpose.

Usage

Auto-Compact must be set to True.
For unattended runs, set askUserQuestionTimeout — an unanswered question then never parks the loop forever.

1. Write your anchor — /ploop:define-mission for a clear goal, /ploop:define-purpose for an ongoing direction.
2. In a fresh session, run /ploop:launch [anchor]. The loop rides the Stop hook's error behavior — whenever the agent stops, the hook blocks the stop and directs it to invoke the advisor.
3. The loop ends on its own when the advisor judges there is nothing left to advise — at which point the agent reads the log and recaps every round. To pause it, run /ploop:off; to pick it back up from where it stopped, run /ploop:on (interrupt with ESC first if a turn is running). off halts the loop quietly and preserves its state; on resumes the loop from that state. on is also a universal wake button for a long-running loop stalled by a mishap — an accidental ESC, an API error, a subscription session limit: it always resumes cleanly, except when the advisor ended the loop itself. Nothing else — mid-run instructions, answered questions, background-task notifications, ESC itself — stops the loop.

Refine

Install: claude plugin install refine@claude-automata
Update: claude plugin update refine@claude-automata

refine is a family of large-scale workflows that eliminate the debt a repository accumulates.

All three skills work the same way — split into regions for parallel analysis, settle findings through a cross-examination assembly, and execute only the highest-ROI plans. Each run is a heavyweight workflow taking hours (3–12h).

• /refine:code [focus] — code architecture optimization. Filters antipatterns through consensus and applies only the highest-ROI refactors.
• /refine:docs [focus] — documentation architecture optimization. Every claim in every non-executable text (markdown, doc systems like openspec, comments and docstrings) is checked against the code and set right. Alignment is the precondition — converging duplicates, deleting dead docs, and keeping docs minimal is the optimum. Code is never modified — code defects are reported.
• /refine:integrity [focus] — integrity-boundary optimization. Hunts the reachable states the existing boundary (types, invariants, error definitions, tests) fails to contain, digs in from "should this be defined as an error?", absorbs them into the boundary, and pins every defined behavior with tests and its rationale in docs and comments.

Leave the focus empty to target the whole codebase. Watch progress with /workflows.

tx - Git Transaction Workflow

Install: claude plugin install tx@claude-automata
Update: claude plugin update tx@claude-automata

tx is a Git workflow that manages change as transactions.

• A transaction is an integrity boundary — everything from open to close is bound into one, and it can only close once verified integral.
• The whole path runs on tx's own skills: /tx:open cuts a tx-* branch off the base branch and seeds the repo, tx:plan·tx:apply·tx:verify drive the change, and /tx:close squash-merges to base behind the docs gate and CI. Guard hooks keep the base branch protected in between.

Prerequisites: uv, Node.js >= 20 (drives the pinned OpenSpec CLI through npx — nothing to install), GitHub CLI (gh).

Usage