Magicblock logo

Magicblock

OrganizationPopular
internet-court
magicblock

Design, implement, and debug MagicBlock applications on Solana. Covers Ephemeral Rollups with delegated state; ER/PER architecture and settlement; private payments and token flows; oracles and randomness; scheduling and temporary authority; security and local validation. Use for MagicBlock product selection, integration, cross-product design, or production troubleshooting.

Overview

Publisherinternet-court
Repositoryinternet-court-skill
Skill namemagicblock
Stars
5.8K
Forks
106
Bundled files
18
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 18 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by internet-court on GitHub. Read the source before you install it.

Installation

Install the Magicblock AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/internet-court/internet-court-skill.git /tmp/internet-court-skill
mkdir -p .claude/skills
cp -r /tmp/internet-court-skill/vendored/magicblock/magicblock-dev .claude/skills/magicblock
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Magicblock in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Magicblock on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Magicblock is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

MagicBlock Development Skill

Pair with the solana-dev skill

Use this skill for MagicBlock-specific concerns: ER/PER, delegation, oracles, Session Keys, cranks, VRF, Magic Actions, eSPL, and private payments. For general Solana or Anchor work such as scaffolding, PDAs, account layouts, SPL tokens, clients, wallets, or LiteSVM/Mollusk testing, also load solana-dev.

Key Concepts

Ephemeral Rollups enable high-performance, low-latency transactions by locking a delegated account on the base layer while an ER clone continues to execute under the account's original program owner. They are useful for gaming, real-time apps, and fast transaction throughput.

Ephemeral Accounts are born, used, and closed only inside an ER. They are useful for temporary high-frequency state, but they never commit to Solana and therefore cannot be the only copy of durable ownership, balances, rewards, or settlement results.

Delegation temporarily assigns the base-layer account to the Delegation Program and clones it into the ER with its original program owner. Normal program ownership, signer, authority, and account constraints still apply on the ER; delegation status is a routing and lifecycle concern, not a new application authorization rule.

Delegation debugging invariant: a properly delegated account looks owned by the delegation program on base, owned by the original program on the ER endpoint returned by router getDelegationStatus, and cloned into the ER with delegated=true.

For the verified SDK v0.16.2 snapshot, use MagicIntentBundleBuilder to schedule commit and commit-and-undelegate intents. Do not use the deprecated free functions commit_accounts and commit_and_undelegate_accounts.

Private Ephemeral Rollups (PER) gate a delegated account inside a TEE-backed validator with an ER-local EphemeralPermission. Delegate only the data account on the base layer, then create, update, and close its permission on the ER with CreateEphemeralPermissionCpi, UpdateEphemeralPermissionCpi, and CloseEphemeralPermissionCpi. Do not create or delegate a separate base-layer permission account.

Magic Actions are base-layer instructions scheduled inside an ER transaction via MagicIntentBundleBuilder.add_post_commit_actions(...). Each attempted base-layer transaction applies its commit and actions atomically. If any BaseAction fails, the committor removes every BaseAction in that affected TransactionStrategy before retrying its remaining commit strategy; actions in other transaction/finalize strategies are outside that removal scope. Observe and reconcile every originally scheduled action: scheduling or eventual commit success alone does not prove that any of them ran.

Commit sponsorship: every delegated account gets 10 free commits to base layer by default. To lift the cap, either re-delegate (refreshes the quota) or attach the validator-scoped magic_fee_vault PDA and a delegated fee payer to the intent bundle. The delegated payer is debited; the fee vault is the validated destination credited with that commit fee.

Lamports top-up: when a delegated account (e.g. a delegated fee payer) needs more lamports on the ER side, use lamportsDelegatedTransferIx from the SDK. The transaction is submitted on base layer — the Ephemeral SPL Token program creates a single-use lamports PDA, funds it, and delegates it so the ER credits the destination.

Ephemeral SPL Token has two surfaces. In the SDK lifecycle model, clients use delegateSpl/transferSpl/undelegateIx/withdrawSpl, and the ER balance appears as a normal SPL token account at the owner's canonical ATA address, so Anchor programs can use plain SPL Token CPI. In the direct-program model, contracts use ephemeral-spl-api and explicitly work with the eATA/global-vault PDAs; do not apply the canonical-ATA model to that raw surface.

Pricing Oracle republishes supported market feeds for Solana/ER consumers. A safe integration verifies the expected feed identity, upstream publish-time freshness, value domain, exponent, checked arithmetic, and user price bounds; successful deserialization alone is not price validation.

Session Keys authorize a temporary signer for constrained application actions. Session validity is separate from SPL token authority: token spending also requires an explicit, bounded token delegate allowance.

Architecture:

┌─────────────────┐     delegate      ┌─────────────────────┐
│   Base Layer    │ ───────────────►  │  Ephemeral Rollup   │
│    (Solana)     │                   │    (MagicBlock)     │
│                 │  ◄───────────────  │                     │
└─────────────────┘    undelegate     └─────────────────────┘
     ~400ms                                  ~10-50ms

Default stack

Programs

Use Anchor with ephemeral-rollups-sdk; native and Pinocchio are also supported.

  • Use the target repo's existing ephemeral-rollups-sdk / Anchor versions unless the task is an explicit upgrade
  • The SDK feature flag selects the Anchor range: anchor for Anchor 1.x programs, or anchor-compat for Anchor >=0.28,<1.0 programs

Required macros:

  • #[ephemeral] on the program module, before #[program] — injects the process_undelegation callback (the delegation program CPIs into it to return the account) and the commit/undelegate intent builders. Commit and undelegation require it; delegation itself does not. Include it on any program that delegates so its accounts can later be undelegated.
  • #[delegate] and #[commit] on the respective delegation/commit account contexts.
  • #[vrf] on a VRF request context and #[vrf_callback] on the VRF callback context — the callback macro authenticates fulfillment. Enable the vrf feature on ephemeral-rollups-sdk. SDK v0.16.2 re-exports VRF, so new Anchor code does not need a direct ephemeral-vrf-sdk dependency. See vrf.md.

Non-Anchor programs: use the ephemeral-rollups-pinocchio crate (delegation, commit, and VRF have Pinocchio equivalents). The engine examples repo ships Anchor and Pinocchio variants of roll-dice; use Pinocchio when the target program is native rather than Anchor.

Versions in this skill are known-good snapshots or compatibility markers. Before changing dependencies, inspect the target repository's manifests, toolchain files, lockfiles, and relevant upstream sources. See resources.md for the dated snapshot and source links.

Connections

  • Base layer connection for initialization and delegation: https://rpc.magicblock.app/devnet or https://rpc.magicblock.app/mainnet
  • Router connection for delegation status: https://devnet-router.magicblock.app/ or https://router.magicblock.app/
  • Ephemeral rollup connection for operations on delegated accounts: use the fqdn returned by router getDelegationStatus

Transaction routing

  • Delegate transactions → Base Layer
  • Operations on delegated accounts → Ephemeral Rollup
  • Undelegate/commit transactions → Ephemeral Rollup

Operating procedure

0. Plan architecture when the design is not fixed

For a new application, integration design, migration, or implementation plan, read architecture-planning.md before writing code. Decide whether MagicBlock is needed, select the smallest product set, map accounts and transaction routing, define settlement and recovery, and choose validation environments. Ask at most three material questions per round; otherwise proceed with explicit assumptions.

1. Classify the operation type

  • Account initialization (base layer)
  • Delegation (base layer)
  • Operations on delegated accounts (ephemeral rollup)
  • Commit state (ephemeral rollup)
  • Undelegation (ephemeral rollup)
  • ER-only Ephemeral Account lifecycle (ephemeral rollup; never commits)
  • Asynchronous service work (VRF callback, crank, queued transfer, or Magic Action)
  • Hosted API transaction construction followed by client signing/submission

2. Pick the right connection

  • Base layer: https://rpc.magicblock.app/devnet or https://rpc.magicblock.app/mainnet
  • Router: https://devnet-router.magicblock.app/ or https://router.magicblock.app/
  • Ephemeral rollup: the fqdn returned by router getDelegationStatus for the account

3. Implement with MagicBlock-specific correctness

For each implementation, record:

  • Which connection to use for each transaction
  • Router getDelegationStatus checks before operations
  • PDA seeds matching between delegate call and account definition
  • Preserving preflight for supported base transactions, and using skipPreflight: true only for an ER path with a known simulation incompatibility (document the reason and inspect execution logs)
  • Waiting for state propagation after delegate/undelegate
  • For Ephemeral SPL Token flows, selecting the deposit and withdrawal builders independently: use the default shuttle withdrawal, or explicitly run undelegateIx, wait for its base commit, and call the legacy withdrawSpl(..., { idempotent: false }); use ephemeral-spl-api exports (not copied bytes or guessed seeds) for direct CPI
  • For oracle flows, feed identity, maximum age, numeric conversion, user limits, and stale-feed behavior
  • For Session Keys, scope, expiry, optional one-time signer lamports top-up, revocation, application- enforced spending limits, and any separate SPL delegate allowance
  • For asynchronous flows, the difference between acceptance/scheduling and completion, plus observation, idempotency, timeout, retry, refund, and reconciliation

For security-sensitive designs, reviews, and implementations, read security.md. Separate protocol guarantees from required integration validation, application policy, and ordinary Solana security. Do not present an application recommendation as a MagicBlock protocol guarantee.

4. Debug live delegation/routing failures

For InvalidWritableAccount, missing private balances, validator mismatch, or "account is delegated but ER rejects it" reports:

  • Start from the exact signature or account pubkey.
  • Query router getDelegationStatus and use its fqdn for ER reads/transactions.
  • Compare base ownership, router status, ER ownership, and recent ER transaction logs.
  • Treat base ownership by the delegation program as expected for a delegated account.
  • See debugging.md for the full runbook.

5. Diagnose possible service-side failures

For unexpected RPC, routing, oracle, or transaction errors that could be service-side:

  • Always fetch current data; do not answer from remembered status. Use the direct JSON API https://status.magicblock.app/api/services as the source of truth.
  • Select the same network the code uses: JSON keys are mainnet and devnet.
  • Match the affected endpoint to the right region/server and service:
    • Regions are asia, europe, usa, and tee.
    • Service IDs are listed in .meta.services; currently er (Ephemeral Rollup), rpc_router, pricing_oracle, and vrf_oracle.
    • Use the server entries under .environments[network].regions[region].servers; for mainnet Asia this includes as.magicblock.app.
  • Interpret .live_status[service]: true = Operational, false = Down, missing/undefined = N/A.
  • Interpret .metrics[service] as downtime minutes per day aligned with .meta.days in UTC.
  • When reporting findings, include the network, region, endpoint, service status, and relevant date range. Distinguish live status from historical downtime.
  • For direct ER RPC endpoints, optionally correlate with JSON-RPC getHealth or getVersion, but do not let a single RPC probe replace the status API.

6. Add appropriate features

  • Cranks for recurring automated transactions
  • VRF for verifiable randomness in games/lotteries
  • Private payments API for private transfers and swaps
  • Pricing Oracle for verified external market data
  • Session Keys for repeated low-friction user actions
  • Ephemeral Accounts for temporary state that is explicitly allowed to disappear

When combining products, read composition-patterns.md and keep authority, token control, oracle validity, ER execution, base settlement, and asynchronous completion as separate guarantees.

7. Deliverables expectations

When you implement changes, provide:

  • Exact files changed + diffs
  • Commands to install/build/test
  • Risk notes for anything touching delegation/signing/state commits

Progressive disclosure (read when needed)

Terminal Self-Heal Proposal

Before the final response, review the entire run against this skill's intended workflow.

Check:

  • expected vs actual user-visible outputs
  • command, script, validation, and artifact results
  • missed coverage or skipped workflow steps
  • stale instructions or brittle wording
  • unexpected script/output mismatches
  • verification gaps
  • fallback or impromptu behavior

If gaps are found, report them with evidence and request explicit approval for a separate maintenance task. Treat installed skill files as read-only during normal execution.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Magicblock AI skill do?

Design, implement, and debug MagicBlock applications on Solana. Covers Ephemeral Rollups with delegated state; ER/PER architecture and settlement; private payments and token flows; oracles and randomness; scheduling and temporary authority; security and local validation. Use for MagicBlock product selection, integration, cross-product design, or production troubleshooting.

Why use Magicblock on TypingMind?

Because you install it once and use it with any model. Magicblock is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Magicblock in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/internet-court/internet-court-skill/tree/main/vendored/magicblock/magicblock-dev. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Magicblock?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Magicblock?

As many as you like. As long as a model supports skills, you can use Magicblock with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Magicblock AI skill free?

It is published on GitHub by internet-court. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇