Claude Code / Codex / AI configs for the expert Solana builder. CLAUDE.md, agents, commands, hooks, rules, skills and settings across Web, Anchor, Pinnochio, Unity, Mobile, security and more.
SaferSkills independently audited solana-ai-kit (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This is the plugin variant of the kit's skill hub. It ships only the skills that travel cleanly in a plugin (go-to-market wrappers + the opt-in add-on catalog). For the full depth — protocol/security/ecosystem skills, .claude/rules/*, and the curated permissions/sandbox policy — use the full install (see "Getting more depth" below).
Source precedence (when multiple skills cover one topic):
.claude/rules/* are law for code style — but rules ship only with the full install (install.sh), not this plugin. If you installed via the plugin, the house code-style rules are not loaded; prefer them when present.solana-dev) are primary for general concepts (Anchor, Pinocchio, testing, clients) — add its marketplace.These load automatically when the plugin is enabled. Commands and skills are namespaced under solana-ai-kit: (e.g. /solana-ai-kit:deploy).
Local wrappers, adapted from sendaifun/solana-new (MIT, telemetry removed):
Plugins cannot carry git submodules, so the 18 external skill packs (the ext submodules) are not bundled here. Two ways to get them:
Several of the upstream skill packs publish their own Claude Code marketplaces. Add the ones you need, then /plugin install the specific skill plugins:
| Domain | Add the marketplace | Then install |
|---|---|---|
| DeFi protocols, infra, data (Jupiter, Raydium, Kamino, perps, oracles, cross-chain…) | /plugin marketplace add sendaifun/skills | the protocol plugins you need (one per skill) |
| AppSec scanning (SAST, SCA, secrets) | /plugin marketplace add ghostsecurity/skills | ghost |
| Security auditing, vulnerability scanning | /plugin marketplace add trailofbits/skills | the audit plugins you need |
| Infrastructure (Workers, Agents SDK, MCP servers) | /plugin marketplace add cloudflare/skills | cloudflare |
For Jupiter, Metaplex, and Helius the official skill repos are the primary sources — check the add-on catalog (skill-registry.json) for their current upstream locations and /plugin marketplace add targets. Route to a marketplace and install the plugin; do not point at an upstream repo's SKILL.md directly.
Run the installer in your project to get everything the plugin can't carry — the 18 external skill submodules, the lazy-loaded .claude/rules files (Rust, Anchor, Pinocchio, TypeScript, .NET code-style law), and the curated permissions + sandbox policy:
curl -fsSL https://raw.githubusercontent.com/solanabr/solana-ai-kit/main/install.sh | bashSee the project README ("External Skill Submodules" and "Install as a Claude Code plugin") for when to pick the plugin vs. the full install — they are complementary, not competing. If you run both in the same project, /solana-ai-kit:doctor warns about duplicate commands/hooks/MCP and advises picking one.
| User asks about... | Skill |
|---|---|
| Idea validation, "what should I build" | idea-sprint/SKILL.md |
| Pitch deck, demo day, investor or grant slides | pitch-deck/SKILL.md |
| Hackathon submission, demo script, track choice | hackathon/SKILL.md |
| Opt-in add-on skills/plugins/MCPs, "should I add X", extra capability not bundled | skill-registry.json |
| Protocol SDK depth (Jupiter, Metaplex, Helius, Raydium…), security audits, infra | add the upstream marketplace (Option A) or do the full install (Option B) |
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.