Private beta

Turn ad-hoc AI use into repeatable procedures

Build a library of reusable AI procedures that compound across your team. Most organizations are stuck at the “clever individual prompts” stage: useful experiments, no reuse, no lasting memory. These Claude plugins fix that.

The very same tools Luca uses in his daily work. Free during the private beta.

Organizations and individuals

luca-ops-kit

Your team's best AI procedures, captured once and reusable forever. Build a self-improving setup where every session makes the next one better, and know-how stops being trapped in individual heads or chat histories.

Most users start here.

Developers

luca-dev-kit

Type "open pr" and walk away. Claude runs triple review, creates the PR, iterates Gemini feedback, and pings you when it's clean. Pre-PR quality gates, autonomous review loops, and pre-commit hooks, all wired together.

Requires GitHub CLI and Gemini Code Assist.

Add this if your team ships code through PRs.

What you can do right after installing

A few highlights. Each plugin ships with more.

luca-ops-kit

/create-skillTry firstTurn any procedure into a reusable skill that improves itself over time.Created /onboard-client with 4 steps, self-review hook enabled.
/reflectEnd every session smarter: surface what worked, what didn't, and what should become a reusable skill or memory.Extracted 2 learnings, updated 1 memory, suggested 1 new skill.
/audit-skillKeep your skill library sharp. Scores quality across 7 dimensions, proposes fixes, and iterates until the bar is met.Your review-agreement skill scored 6.2 → applied 3 fixes → re-scored 9.1.
/build-work-contextTell Claude about your role and company once; it remembers forever. No more repeating yourself each session.Saved profile: senior PM, fintech, focus on compliance workflows.
/dreamMine patterns across sessions. Consolidates what /reflect has learned into cross-cutting insights and skill ideas.Found 3 recurring patterns, suggested 2 new skills.

luca-dev-kit

/open-prTry firstSay "open pr" and walk away. Claude reviews, pushes, creates the PR, and hands off to the Gemini review loop.Triple review passed, PR #142 created, review-loop started.
/review-loopFully autonomous: polls Gemini, classifies comments, applies fixes, and re-triggers review until the PR is clean.3 rounds, 7 comments resolved, 0 remaining.
/triple-reviewCatch issues before reviewers do. Three parallel lenses: principles, recurring bug patterns, and structural integrity.Found 1 principle violation, 0 bugs, 1 structural suggestion.
/specs-adherence-reviewNever ship code that breaks your own rules. Checks every change against your CLAUDE.md and AGENTS.md.8 rules checked, 1 violation found and auto-fixed.

Before you install

I'm sharing this with a small group of people I trust. The terms below look formal because I want us to be properly covered, but this is early access between friends. If anything feels unclear, just reach out.

The terms below are essentially a “use at your own risk” agreement. They say: the plugins are for business use, AI outputs can be wrong so review them before acting, I have no control over Anthropic or Claude Code and can't be held responsible if they change, and my liability is capped at fees you paid in the preceding 12 months (zero for early access). Nothing unusual for software of this kind. This summary is informational; the full terms govern.

These plugins (luca-ops-kit and luca-dev-kit) are for business use only, intended exclusively for businesses and professionals acting in the course of a trade or profession. By using them, you represent that you are not a consumer under Italian law (Article 3, Codice del Consumo). They are designed exclusively for use with Claude (Anthropic); any adaptation to other AI providers is unsupported and at the user's sole risk. They are provided "as-is", without warranty of any kind, express or implied. The plugins depend on third-party services (including Anthropic and Claude Code) that Luca Dellanna does not control. Changes to those services, including model deprecations, pricing changes, or discontinuation, are not Luca Dellanna's responsibility. The plugins run locally on your machine; Luca Dellanna does not collect or process your data. Outputs are generated by Anthropic Claude; the plugins orchestrate prompts but do not themselves produce content. AI-generated outputs may contain errors, fabrications, or hallucinations, and may appear authoritative but be factually wrong. You are solely responsible for reviewing all outputs before acting on them. These plugins are not a substitute for professional legal, financial, medical, or other expert advice; do not use them for legal filings, medical decisions, financial transactions, regulated activities, or high-risk AI applications (EU AI Act, Reg. 2024/1689) without independent expert review. To the maximum extent permitted by applicable law, and in any case limited to fees paid in the 12 months preceding the claim, Luca Dellanna accepts no liability for any loss, damage, or consequence, whether direct or indirect, arising from use of these plugins or reliance on their outputs. This includes losses caused by third-party service changes (Anthropic, Claude Code, etc.). Nothing in this disclaimer excludes liability that cannot be excluded under applicable law (including liability for fraud, gross negligence, or death or personal injury caused by negligence). Full terms will be published at luca-dellanna.com before commercial launch. This disclaimer is governed by Italian law; any disputes shall be resolved in the Italian courts.

Read full draft terms on GitHub

Full name

Work email

Used solely to record your acceptance of the terms.

On behalf of my organization, I confirm I have read and accept the Terms of Use (v1.0, dated 2026-05-10), and that I have authority to bind the company to these Terms.

Specific approval of onerous clauses (Art. 1341 c.c.)

Pursuant to Articles 1341 and 1342 of the Italian Civil Code, I specifically approve the following clauses: 4 (No warranty), 5 (AI output disclaimer), 7 (Limitation of liability), 9 (Acceptable use), 11 (Customer indemnification), 12 (Fees and refunds), 13 (Term and termination), and 19 (Governing law and exclusive forum).

Turn ad-hoc AI use into repeatable procedures

The very same tools Luca uses in his daily work. Free during the private beta.

luca-dev-kit

Requires GitHub CLI and Gemini Code Assist.

Add this if your team ships code through PRs.