v0.2.1

Changed

• Described the skill by what it does, not why: "a generic skill that runs multiple roborev reviewers in a loop and tracks every finding in a private ledger," replacing "keeps a long loop honest," in the README, the skill's triggering description, and the plugin manifests.
• Fixed a stale 0.1.0 status badge in the README.

Full diff: v0.2.0...v0.2.1

---

Authored by Jesse Robbins (@jesserobbins)

v0.2.0

Changed

• Rewrote the README to lead with who this skill is for, what it does, and how it works, with a concrete example pulled from the worked example.
• Replaced "discipline" throughout the README, the skill (including its triggering description), the changelog, and the plugin manifests with plainer language: this is a skill that enhances roborev, not a discipline layered on top of it.
• Renamed the skill's "Type discipline" section to "Type rules".

Full diff: v0.1.0...v0.2.0

---

Authored by Jesse Robbins (@jesserobbins)

v0.1.0

First release out of alpha. The two-step marketplace install has been run end to end.

Added

• A scope note in the skill that states plainly what the ledger reads: only roborev's own review findings, tied to a specific job ID, never an outsider-authored feed, queue, or text channel.
• An Author section in the README.

Changed

• Rewrote all prose in the README, changelog, worked example, and skill in Simplified Technical English: short sentences, active voice, no banned modals, no filler.
• Removed the alpha status note and the "install not yet verified" known limitation, now that the install has been verified end to end.

Full diff: v0.0.2-alpha...v0.1.0

---

Authored by Jesse Robbins (@jesserobbins)

v0.0.2-alpha — Robo Refine and Polish

Changed

• Rename the standalone skill to roborev-refine-and-polish.
• Use Iteration consistently in the skill and worked example.
• Record per-iteration plans and results in a table.
• Update plugin metadata and documentation.

v0.0.1-alpha — Robo Refine and Polish

First alpha. Packages the Robo Refine and Polish discipline as a standalone, self-installing Claude Code plugin.

The discipline is real and battle-tested in practice; the packaging is new. The two-step marketplace install follows the documented plugin flow and matches working plugins, but hasn't been run end-to-end by a fresh user yet — so expect rough edges in the plumbing, not the method.

Install

/plugin marketplace add jesserobbins/refine-and-polish
/plugin install refine-and-polish@refine-and-polish

Added

• The refine-and-polish skill — the full discipline for tracking a multi-iteration roborev refine loop in a private ledger: typed findings (NEW / REGRESSION / REPEAT / LOOP), the deliberate-pushback list, reviewer-coverage caveats, an explicit convergence criterion and iteration budget, and the honest "budget stop" for an unbounded Low-severity tail.
• Plugin packaging: plugin.json and a self-contained marketplace.json, so the repo installs as its own marketplace.
• A worked example — a real three-iteration loop with a filled-in ledger, a pushback entry, and an honest convergence caption.
• README quickstart and an alpha-status note.

Known limitations

• The two-step marketplace install matches the documented plugin flow and working plugins, but has not been run end-to-end by a fresh user.
• Requires the roborev CLI; this is a discipline layered on top of it, not a standalone tool.
• Assumes subscription-backed reviewers; the "reviewer unavailable" handling deliberately refuses API-key fallback.

Requires roborev.