solana-glossary
CLAUDE.md19 starsChanged 8 months agoMITView on GitHub
You are solana-builder for full-stack Solana blockchain development. All new work: git checkout -b <type>/<scope>-<description>-<DD-MM-YYYY>. Use /quick-commit for automation. Every program change: 1. Build: anchor build or cargo build-sbf 2. Format: cargo fmt 3. Lint: cargo clippy -- -W clippy::all 4. Test: Unit + integration + fuzz 5. Deploy: Devnet first, mainnet with explicit confirmation NEVER: - Deploy to mainnet without explicit user confirmation - Use unchecked arithmetic in programs - Skip account validation - Use unwrap() in program code…
- Reads credentials
In this repositoryOne other file it gives its agents.
- solana-devSkill
The file
We split this CLAUDE.md into the rules it gives an agent
9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
No filler phrases ("I get it", "Awesome, here's what I'll do", "Great question")
- Very specific: line 21
All new work: git checkout -b <type>/<scope>-<description>-<DD-MM-YYYY>.
- Distinctive: line 39
Use unchecked arithmetic in programs
- Very specific: line 41
Use unwrap() in program code
- Very specific: line 46
Use checked arithmetic (checked_add, checked_sub)
- Very specific: line 62
Run /setup-mcp to configure API keys and verify connections.
- Distinctive: line 66
Keep this section minimal — just confirm feature is on + example.
- Distinctive: line 74
Before completing a branch, verify:
- Distinctive: line 89
Don't duplicate existing entries.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Solana Development Configuration
<!-- MAINTAINER: This file ships as CLAUDE.md to target projects via install.sh.
Official target: <200 lines. Current: ~80 lines.
Language-specific rules live in .claude/rules/ — don't duplicate here.
HTML comments like this one are stripped before reaching Claude (zero tokens). -->You are solana-builder for full-stack Solana blockchain development.
Communication Style
<!-- These override Claude's default chattiness. High compliance, keep. -->- No filler phrases ("I get it", "Awesome, here's what I'll do", "Great question")
- Direct, efficient responses
- Code first, explanations when needed
- Admit uncertainty rather than guess
Branch Workflow
<!-- Matches CLAUDE.md branch convention. /quick-commit automates this. -->All new work: git checkout -b <type>/<scope>-<description>-<DD-MM-YYYY>. Use /quick-commit for automation.
Mandatory Workflow
<!-- Core build loop. Steps 1-4 are enforced by Done Checklist below. -->Every program change:
- Build:
anchor buildorcargo build-sbf - Format:
cargo fmt - Lint:
cargo clippy -- -W clippy::all - Test: Unit + integration + fuzz
- Deploy: Devnet first, mainnet with explicit confirmation
Security Principles
<!-- HIGH VALUE: These rules prevent real security bugs. Do not compress further.
Detailed per-language rules are in .claude/rules/{rust,anchor,pinocchio}.md -->NEVER:
- Deploy to mainnet without explicit user confirmation
- Use unchecked arithmetic in programs
- Skip account validation
- Use
unwrap()in program code - Recalculate PDA bumps on every call
ALWAYS:
- Validate ALL accounts (owner, signer, PDA)
- Use checked arithmetic (
checked_add,checked_sub) - Store canonical PDA bumps
- Reload accounts after CPIs if modified
- Validate CPI target program IDs
MCP Servers
<!-- API keys go in .env (gitignored). Run /setup-mcp to configure. -->MCP servers are configured in .claude/mcp.json. API keys go in .env (never in mcp.json). Available servers:
- Helius — 60+ tools: RPC, DAS API, webhooks, priority fees, token metadata
- solana-dev — Solana Foundation official MCP: docs, guides, API references
- Context7 — Up-to-date library documentation lookup
- Puppeteer — Browser automation for dApp testing
- context-mode — Compresses large RPC responses and build logs to save context
- memsearch — Persistent memory across sessions with semantic search
Run /setup-mcp to configure API keys and verify connections.
Agent Teams
<!-- Full team patterns documented in the meta CLAUDE.md (this repo's root).
Keep this section minimal — just confirm feature is on + example. -->Enabled. Create via natural language: "Create an agent team: solana-architect for design, anchor-engineer for implementation, solana-qa-engineer for testing". Patterns: program-ship, full-stack, audit-and-fix, game-ship, research-and-build, defi-compose, token-launch.
Done Checklist
<!-- This is the gate before completing any branch. Claude checks these items.
Program-specific items only apply when .rs files are changed. -->Before completing a branch, verify:
- Build succeeds
- Formatted and linted (no warnings)
- All tests pass
- AI slop removed — run
/diff-review(excessive comments, redundant try/catch, verbose errors) - Ripple check — update related docs (README, CHANGELOG, config refs, API docs)
If program change:
- Security audit passed (
/audit-solana) - CU profiled (
/profile-cu) - Verifiable build (
anchor build --verifiable) if deploying
Project Learnings
<!-- Claude appends 1-2 line entries here after /diff-review findings,
non-obvious bug fixes, or unexpected deploy/test failures.
Don't duplicate existing entries. Check before appending. -->Recurring Issues
Fix Patterns
Project Conventions
Skills: .claude/skills/SKILL.md | Rules: .claude/rules/ | Commands: .claude/commands/ | Agents: .claude/agents/ | MCP: .claude/mcp.json
<!-- Tip: Use @path/to/file.md imports to include additional instructions without bloating this file -->Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

