Pular para o conteúdo
TakoAPI
Voltar às skills

aip-identity

por the-nexus-guard

Coding Agents & IDEs
1.1kDownloads informados pelo clawskills.sh
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

Digital identity, authentication, and trust for AI agents.

Descrição

O que esta skill faz

Cryptographic identity and trust infrastructure for AI agents, built on the Agent Identity Protocol. Agents get a decentralized identifier backed by an Ed25519 keypair, enabling verifiable authentication, signed content, end-to-end encrypted messaging, and scoped trust vouching between agents. Keys can be rotated without losing the existing identity.

Provides verifiable agent identity without blockchain, tokens, or staking — just Ed25519 cryptography with portable DIDs and time-decaying trust scores.

Quando usá-la

  • Verifying another agent's identity before granting access to a resource
  • Signing a skill file to prove authorship and detect tampering
  • Sending encrypted messages between two agents without the server reading them
  • Building a trust network by vouching for known agents with scoped permission levels
  • Rotating signing keys after a credential exposure without losing the existing DID

Exemplo de fluxo de trabalho

ENTRADA

User asks: sign my skill file and send the signed proof to another agent securely

AGENTE
  1. 1

    Register with --secure flag to generate a local Ed25519 keypair and DID

  2. 2

    Run sign --file my_skill.py to produce an Ed25519 signature over the file hash

  3. 3

    Look up the recipient's DID using verify --username TargetAgent

  4. 4

    Send the signature via message --recipient-did <did> --text <proof>

  5. 5

    Recipient runs messages to auto-decrypt and read the signed proof from inbox

SAÍDA

Skill file is cryptographically signed with verifiable authorship, and the proof is delivered as an encrypted agent-to-agent message

Instalação

clawhub install the-nexus-guard/aip-identity

Execute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.

Links