Improve the current branch by running /code-improver:improve — a dynamic workflow that\nloops a PR reviewer and a fixer subagent over the branch's changes until a review\nreports zero critical/major findings. The loop, its ledger, and its guards live in the\nworkflow; this skill derives the scope from the branch diff and relays the outcome.
The user provided: $ARGUMENTS (if empty, take base branch and preferences from the\nconversation).
git rev-parse --show-toplevel. Fail loudly outside a repository.git symbolic-ref refs/remotes/origin/HEAD → its short name, falling back to\nmain). Refuse to run when the current branch IS the base — there is no diff to\nimprove.git diff --name-only <base>...HEAD. If empty, say so and stop.<dir>/** (** at the repo root only if files at the\nroot changed), then deduplicate and drop globs covered by another.The loop is the dynamic workflow workflows/improve.js in this plugin. Launch it by\npath: scriptPath takes a resolved absolute path, and the Workflow tool's name resolves\nbuilt-in and project workflows, so a marketplace-installed one may not answer to\ncode-improver:improve. Try in order, first hit wins — the home directories come before\n. so an installed copy beats a checkout of this marketplace:
Bash: ls -d -- \"${CLAUDE_PLUGIN_ROOT}/workflows/improve.js\"Bash: ls -d -- \"${CODEX_PLUGIN_ROOT}/workflows/improve.js\" (if that variable is set instead)Bash: find ~/.claude ~/.codex . -maxdepth 7 -path '*/code-improver/workflows/improve.js' -print -quit 2>/dev/nullUse the path exactly as printed. Its plugin directory — the path with\n/workflows/improve.js removed — is pluginRoot. If all three come back empty, try\n{name: \"code-improver:improve\"} once; if that is unavailable too, stop and say the loop\ncould not be located. Do not assemble a path by hand and do not improvise the loop.
Run it with the Workflow tool, {scriptPath: \"<the path from step 2>\", args: {...}}:
{\n \"target\": \"<repo root>\",\n \"reviewer\": {\n \"kind\": \"skill\",\n \"name\": \"pr-review-toolkit:review-pr\",\n \"notes\": \"Review the working tree's changes against <base> as a pull request: correctness, tests, error handling, and the review dimensions the skill prescribes.\"\n },\n \"scope\": [\"<derived-dir-glob>/**\"],\n \"pluginRoot\": \"<the plugin directory from step 2>\",\n \"maxRounds\": 5\n}\nreviewer — the default above requires the pr-review-toolkit plugin. When the user\nnames a different PR reviewer (skill or agent), use it, with kind set accordingly.maxRounds only if the user asked for a different cap.pluginRoot lets the run find its metrics collector; omit the key only if step 2 fell\nthrough to the workflow name — the workflow then searches for itself.finalize defaults are right for PRs: no version bump unless the branch sits inside a\nplugin, narration strip and docs pass on.decision only on continuation (below).The loop's baseline snapshot is the tree at loop start — its scope guard protects the\nbranch's uncommitted work; the PR's own commits are what the reviewer reviews.
\nThe workflow runs in the background and needs no babysitting: it reviews, fixes,\nre-reviews, checks scope after every fix round, and can only complete on a clean review.\nIt never commits; all changes stay in the working tree.
\nIf the Workflow tool is unavailable or denied, stop and say so. Do not improvise the\nloop inline with direct edits — the ledger, scope guard, and escalation guarantees live\nin the workflow, and an inline imitation has none of them.
\nIf the result is halted: \"reviewer-unavailable\", relay it and stop. The reviewer\nis not installed in this session; tell the user which plugin provides it (the default\nneeds pr-review-toolkit) and re-run after installing. Do not review the branch\nyourself.
Do not end your turn while the loop is running. The Workflow tool returns a task id\nimmediately; the result comes later. In an interactive session the completion\nnotification re-invokes you — wait for it. In a non-interactive run (scripted, CI, eval)\nthere is no later turn: stopping abandons the loop mid-round, so after launching, poll\nthe task (TaskOutput with the returned task id, or sleep-and-recheck) until it completes,\nthen relay the result. A session that answers \"the loop is running, I'll report later\"\nhas lost the run.
\nThe workflow returns a structured result. Report it honestly — the distinctions matter:
\nconverged: true — the last action was a review with zero critical/major findings.\nReport rounds used, remaining minor findings (open_minor_count), and the artifact\npaths (ledger_path, metrics).capped: true — the fix budget ran out and the FINAL review still found blocking\nissues. Say plainly: capped, NOT converged, and list open_blocking. Do not\npresent this as success.escalation — the loop detected it was not converging (recurring findings,\nnon-decreasing counts, or a fix relocating a problem). Relay the escalation message and\nfinding ids to the user: this needs a design decision, not more rounds.halted — a guard fired (scope violation, unregistered new files, a dead or\nunavailable reviewer, or a finalize pass whose own edits failed the check that follows\nit). Relay the paths in violations/new_untracked_files, the sites in\nfinalize_regressions, and the notes.notes always travel with the result — surface them; they include loud warnings\nsuch as \"a git repository was initialized\".The loop stops on escalation by design. When the user decides, start a fresh run with\nthe same args plus:
\n{ \"decision\": \"<the user's ruling, verbatim>\" }\nThe new run reloads the on-disk ledger, so every finding, rejection, and verdict carries\nover — rounds restart, re-derivation does not.
\nTo stop a running loop, stop the workflow task (TaskStop); the ledger on disk is current\nto the last round and a re-run resumes from it.
\nskill-improver entry — it wires the right reviewer