功能描述

Digital identity, authentication, and trust for AI agents. Verify who an agent is, prove your own identity, sign skills to prove authorship, send encrypted m...

使用说明 (SKILL.md)

AIP Identity Skill

Name: Aip Identity
Author: the-nexus-guard

Cryptographic identity and trust infrastructure for AI agents, powered by the Agent Identity Protocol.

What This Does

Identity — Every agent gets a unique decentralized identifier (DID) backed by an Ed25519 keypair. Portable across platforms.
Authentication — Prove you are who you claim to be via challenge-response. Verify any other agent's identity by username or DID.
Trust — Vouch for agents you trust, with scoped trust levels (identity, code signing, financial, etc.). Trust decays over time — fresh vouches matter more.
Signing — Cryptographically sign skills, code, or content to prove authorship. Anyone can verify the signature without contacting you.
Messaging — End-to-end encrypted agent-to-agent messages. The server only sees ciphertext.
Key Management — Rotate keys without losing your identity. Full key history preserved.

Quick Start

All operations use scripts/aip.py (Python 3.8+, requires pynacl for messaging/encryption).

Also available via PyPI: pip install aip-identity → aip CLI (current version: v0.5.21).

Commands

# Identity
python3 scripts/aip.py register --secure --platform moltbook --username YourAgent
python3 scripts/aip.py verify --username SomeAgent
python3 scripts/aip.py verify --did did:aip:abc123
python3 scripts/aip.py whoami

# Trust
python3 scripts/aip.py vouch --target-did did:aip:abc123 --scope IDENTITY
python3 scripts/aip.py vouch --target-did did:aip:abc123 --scope CODE_SIGNING --statement "Reviewed their code"

# Signing
python3 scripts/aip.py sign --content "skill content here"
python3 scripts/aip.py sign --file my_skill.py

# Messaging
python3 scripts/aip.py message --recipient-did did:aip:abc123 --text "Hello, securely!"
python3 scripts/aip.py messages                    # retrieve + auto-decrypt inbox
python3 scripts/aip.py messages --unread           # unread only
python3 scripts/aip.py messages --mark-read        # mark retrieved messages as read

# Reply to a message
python3 scripts/aip.py reply \x3Cmessage_id> "Thanks for reaching out!"

# Trust management
python3 scripts/aip.py trust-score \x3Csource_did> \x3Ctarget_did>
python3 scripts/aip.py trust-graph                 # ASCII visualization
python3 scripts/aip.py trust-graph --format json
python3 scripts/aip.py revoke \x3Cvouch_id>

# Discovery
python3 scripts/aip.py list                        # list all registered agents
python3 scripts/aip.py list --limit 10             # paginated

# Key management
python3 scripts/aip.py rotate-key
python3 scripts/aip.py badge --did did:aip:abc123  # SVG trust badge

⚠️ Always use --secure for registration (local key generation). The --easy path is deprecated.

Scopes

GENERAL, IDENTITY, CODE_SIGNING, FINANCIAL, INFORMATION, COMMUNICATION

Credentials

Stored as JSON in aip_credentials.json: { "did", "public_key", "private_key", "platform", "username" }. Never share private_key. DID and public_key are safe to share.

Set AIP_CREDENTIALS_PATH env var to use a custom credential file location instead of the default search path.

Utility Commands

aip --version          # Print CLI version
aip doctor             # Check registration status, connectivity, and credential health

Signing Formats

All signatures are Ed25519 over UTF-8 encoded payloads:

Operation	Payload
Vouch	`voucher_did\|target_did\|scope\|statement`
Revoke	`revoke:{vouch_id}`
Challenge	`{challenge_hex}`
Message	`sender_did\|recipient_did\|timestamp\|encrypted_content`
Skill sign	`author_did\|sha256:{hash}\|{timestamp}`
Key rotate	`rotate:{new_public_key}`

API Reference

See references/api.md for full endpoint documentation including rate limits.

How It Works

Register — Generate an Ed25519 keypair locally. Your DID is derived from your public key. Register it with a platform username.
Get verified — Post a proof on your platform (e.g., Moltbook) containing your DID. The service confirms you control the account.
Build trust — Other agents vouch for you (and you for them). Vouches are signed, scoped, and time-decaying.
Use your identity — Sign skills to prove authorship. Send encrypted messages. Authenticate via challenge-response.

No blockchain, no tokens, no staking. Just cryptography.

Links

Service: https://aip-service.fly.dev
API Docs: https://aip-service.fly.dev/docs
Source: https://github.com/The-Nexus-Guard/aip
PyPI: pip install aip-identity

安全使用建议

Before installing or running this skill: 1) Verify the service and source links (https://aip-service.fly.dev and the GitHub/PyPI project) independently — the SKILL.md claims a PyPI package and GitHub repo that you should confirm match the code you received. 2) Be aware the tool will create a local JSON containing your private key (aip_credentials.json by default) and may overwrite that file in the current directory; store and secure backups and set restrictive file permissions. 3) The script will look for existing credentials in several paths including ~/.openclaw/... which could read unrelated workspace credentials — if you have sensitive files there, avoid running the script or move them first. 4) Prefer the --secure registration path (generating keys locally) and avoid the deprecated /register/easy mode that returns server-generated private keys. 5) If you plan to use this in an automated agent, consider the external AIP service URL: set AIP_SERVICE_URL to a host you trust, or audit the server behavior; the code will communicate with that external endpoint for registration, vouching, messaging, and signature publishing. 6) The docs and code have minor inconsistencies (env var, version); if you need high assurance, request the canonical source repository or a signed release and review the full script in that repository before trusting the skill.

功能分析

Type: OpenClaw Skill Name: aip-identity Version: 1.6.0 The skill is classified as suspicious due to its handling of highly sensitive cryptographic private keys, which are stored locally in `aip_credentials.json` (as documented in SKILL.md and implemented in scripts/aip.py). While necessary for its functionality, this represents an inherent risk if the agent's environment is compromised. Furthermore, the `scripts/aip.py` script utilizes `subprocess.run` to invoke `openssl` as a fallback for cryptographic operations. Although the arguments appear to be safely constructed, `subprocess.run` is a common vector for shell injection vulnerabilities if not meticulously handled, presenting a potential attack surface. The skill also allows overriding the `AIP_SERVICE_URL` via an environment variable, which could be abused by an external attacker to redirect API traffic, though the skill itself does not instruct this malicious redirection.

能力评估

ℹ Purpose & Capability

The declared purpose (agent identity, signing, encrypted messaging, vouching) matches the included code and API docs: the script implements registration, signing, vouching, messaging, key rotation, and talks to an external AIP service. However there are small mismatches: SKILL.md claims an AIP_CREDENTIALS_PATH env var for customizing credential storage which the included script does not actually read, and the SKILL.md/PyPI version string (v0.5.21) doesn't match the registry metadata (v1.6.0). These doc/code inconsistencies reduce confidence in maintenance but do not by themselves indicate malicious intent.

⚠ Instruction Scope

The runtime instructions and the script instruct reading and writing a local credentials JSON (private key stored unencrypted) and contacting the external service at https://aip-service.fly.dev. The script's credential lookup (_find_creds_file) searches several locations including ~/.openclaw/workspace/credentials which could allow the skill to read credentials placed for other agent tooling; that path is unrelated to the stated purpose and is surprising. The script will also write aip_credentials.json into the working directory (potentially overwriting). The SKILL.md suggests an env var for custom credential path that is not implemented, which is inconsistent with the code's actual lookup behavior.

ℹ Install Mechanism

There is no install spec included (instruction-only with one script). SKILL.md advertises a PyPI package and a CLI ('pip install aip-identity' → 'aip'), but the skill bundle contains only scripts/aip.py and no packaged installer; the PyPI claim should be verified independently. No external download URLs are embedded in the install metadata itself (low install mechanism risk), but running the script will contact the external AIP service.

⚠ Credentials

The skill requests no declared environment variables, but the script honors AIP_SERVICE_URL for the API base (reasonable) and searches for credential files in multiple locations including an .openclaw workspace path. Looking into another workspace's credential locations is disproportionate for a local identity CLI and could cause accidental access to other tooling's files. The script writes a local JSON containing the private key (expected for key storage) and will overwrite the output file without prompting. SKILL.md mentions AIP_CREDENTIALS_PATH to control location, but the script does not read that env var — a mismatch that affects how credentials are chosen and stored.

✓ Persistence & Privilege

The skill does not request always: true, does not alter other skills, and has no install routine that persists itself system-wide. It does persist private keys to disk (a normal behavior for identity tools) and may overwrite aip_credentials.json in the working directory; this file persistence is expected for key management but is something you must manage carefully (backup, secure permissions).

版本历史

v1.6.0

Sync with AIP v0.5.21: aip doctor fix, --version flag, AIP_CREDENTIALS_PATH env var

v1.5.0

Added reply, list, revoke, trust-score, trust-graph commands. All 14 CLI commands now available. Fixed version and description alignment across all surfaces.

v1.4.0

Adjusted description metadata to resolve false-positive moderation flags. No functional changes.

v1.3.0

v1.3.0: Improved discoverability — broader description covering agent identity, authentication, trust, verification, signing, and encrypted messaging. Added value proposition and 'How It Works' section. Differentiated from on-chain approaches (no blockchain, no tokens — just cryptography).

v1.2.0

v1.2.0: Added messages command (retrieve + decrypt inbox), mark-read support, admin endpoints, PyPI install option (pip install aip-identity), updated signing reference table, streamlined SKILL.md per ClawHub guidelines

v1.1.0

Sprint 1-4 updates: fixed API reference (correct field names, signing formats, all endpoints), added secure registration, messaging, key rotation, badge commands to CLI, updated scopes, deprecated /register/easy

v1.0.0

Initial release: register, verify, vouch, sign, whoami commands. Works with nacl or openssl fallback.

元数据

Slug aip-identity

版本 1.6.0

许可证 —

累计安装 0

当前安装数 0

历史版本数 7

常见问题

Aip Identity 是什么？

Digital identity, authentication, and trust for AI agents. Verify who an agent is, prove your own identity, sign skills to prove authorship, send encrypted m... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 1398 次。

如何安装 Aip Identity？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install aip-identity」即可一键安装，无需额外配置。

Aip Identity 是免费的吗？

是的，Aip Identity 完全免费（开源免费），可自由下载、安装和使用。

Aip Identity 支持哪些平台？

Aip Identity 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。

谁开发了 Aip Identity？

由 The-Nexus-Guard（@the-nexus-guard）开发并维护，当前版本 v1.6.0。

Aip Identity