ACTA Agent Skills

Official Agent Skills for building with ACTA, the non-custodial Verifiable Credentials infrastructure on the Stellar blockchain.

A skill is reusable context that teaches an AI coding agent (Claude Code, Cursor, Copilot, Windsurf, Gemini, and others) how to do something correctly. Install one and your agent stops guessing about ACTA: package names, base URLs, the prepare/submit flow, did:stellar issuer identity, error codes, and the security model come from the docs, not from stale training data.

Browse on skills.sh: skills.sh/acta-team/skills · Docs: docs.acta.build/en/agent-skills

Skills in this repo

Skill	Page	What it does
acta	skills.sh	End-to-end ACTA integration: issue, verify, and revoke W3C Verifiable Credentials with the @acta-team/credentials SDK or the REST API; create single-tenant vaults; manage issuer allow/deny; resolve did:stellar; handle errors; understand the security model.

Install

Install into any supported agent with the skills CLI:

# Install every skill in this repo
npx skills add ACTA-Team/skills

# Or install just the acta skill
npx skills add ACTA-Team/skills/acta

The CLI writes the skill into your agent's skills directory (for example .claude/skills/ for Claude Code) and the agent picks it up automatically. See the skills.sh docs for per-agent details.

Manual install

Each skill is a self-contained folder with a SKILL.md. You can also copy the acta/ folder directly into your agent's skills directory.

What the acta skill covers

• The @acta-team/credentials React SDK: ActaConfig, useCredential, useVault, useVaultRead, useActaClient, and ActaClient.
• The REST API: /contracts/* endpoints, the X-ACTA-Key header, the prepare/submit XDR flow, ownership rules, rate limits, and idempotency.
• did:stellar issuer and holder identity, resolution, and issuer verification.
• Vaults: single-tenant, deterministic addressing, sponsored vaults, and push.
• Error handling: HTTP codes, validation, issuer-DID errors, contract errors, and the mainnet USDC fee failures.
• The security and data model: who signs what, what is on-chain vs encrypted.

The skill is structured for progressive disclosure. acta/SKILL.md is the entry point; deeper material lives in acta/references/, with runnable code in acta/examples/.

Source of truth

These skills are distilled from the official documentation at docs.acta.build, which is also queryable live through the @acta-team/docs-mcp MCP server:

{
  "mcpServers": {
    "acta-docs": { "command": "npx", "args": ["-y", "@acta-team/docs-mcp"] }
  }
}

When a detail in a skill conflicts with the live docs, the live docs win.

Safety and scope

These skills are documentation only. They are passive context an agent reads while writing code:

• No execution. No runnable automation, no install hooks, no network calls. Files under examples/ are illustrative snippets for a developer to adapt.
• No secrets. Nothing here asks for, stores, logs, or transmits a secret key, seed phrase, or API key. API keys come from the integrator's own environment.
• No custody, no signing. ACTA is non-custodial: its API returns an unsigned transaction (XDR), the end user's own Stellar wallet signs it locally, and the signed result is submitted. Neither ACTA nor these skills ever hold a private key or can authorize a transfer.
• No payments initiated. The on-chain issuance fee (5 XLM on testnet, 1 USDC on mainnet) is charged by the vault contract and paid by the integrator's own issuer wallet under that wallet's own signature. The skill only documents it.

Because the subject matter is credential issuance on a blockchain, the content necessarily references wallet signing and on-chain fees. That is documented subject matter, not a capability these skills hold.

Contributing

Issues and pull requests are welcome. Keep skills accurate to the current ACTA release, favor concise SKILL.md files with detail pushed into references/, and review any executable content before you install it.

License

MIT