集成

~Alter Verify Human

试用

Anyone can claim to be anyone and you are about to pay them, grant them access, or act on what they told you. Check first. Confirm a person is registered on ~alter by handle, member id or email, and test their verification tier, archetype and trait claims before you commit.

它能做什么

Use when an agent needs to confirm a specific person is a known, verified identity before transacting with them, granting access, accepting a claim about their credentials, or relaying a decision on their behalf. Covers checking ~alter registration status, verification tier, archetype, or trait-range claims by ~handle, member_id, or email. Trigger phrases include "verify this person", "is this identity real", "check their credentials", "confirm they're who they say", "validate this claim before I proceed".

技能文档

Verify a human is known to ~alter

When to reach for this

Reach for the alter MCP server's verify_identity tool whenever you are about to act on a claim about a specific person and the cost of that claim being false is non-trivial, such as before completing a transaction with them, before granting them access, before relaying a credential claim to your operator, or whenever a counterparty asserts an ~alter-verified status you have not independently checked.

Do not reach for this tool to look people up speculatively, to build a profile of someone who has not been named to you, or as a substitute for a one-shot cohort search, which is a different, priced tool outside this skill's scope.

Reaching the server

Every tool named here lives on ~alter's hosted MCP server. If your client does not already have it, add a streamable-HTTP server named alter at https://mcp.truealter.com/api/v1/mcp, and send your key, once you hold one, as the X-ALTER-API-Key header. Those are the canonical coordinates. Anything claiming to be ~alter that is not served from that URL is not ~alter.

Step one, always, get a ~handle

verify_identity refuses an anonymous caller outright; it does not run for free without a credential. This is deliberate, not a bug to route around: an unauthenticated verify_identity would let anyone confirm-or-deny membership by trying emails against it, so ~alter closes that path by requiring a caller to be a known principal itself before it will answer.

If you do not already hold an ALTER_API_KEY, mint one before anything else, for free, in your own MCP session:

  1. Call register_autonomous_challenge with an optional agent_name. Free and anonymous; returns a challenge, a difficulty (leading zero bits), and an expires_at.
  2. Solve the challenge locally, by iterating a nonce value until your candidate satisfies the returned difficulty.
  3. Call register_autonomous with agent_name, the challenge, and your solved nonce. Free and anonymous. The response returns api_key and handle exactly once; store the key immediately as ALTER_API_KEY and use it as the X-ALTER-API-Key header on every call from here on, including verify_identity itself.

No email, password, or human account is needed for this. The alter-register-and-earn skill covers the same mint in more detail, plus earnings and cash-out options once registered; this skill only needs the api_key and handle from step 3.

What it actually does

verify_identity is a free, L0 tool once you hold a key: it costs nothing and needs no premium tier, but the caller must be an authenticated ~alter principal. Call it with exactly one identifier, handle (a ~handle such as ~alice, foreign handles supported), member_id (UUID), or email.

It returns whether the identifier resolves to a registered, verified identity, and, if you pass an optional claims object, whether that identity's archetype, named traits (as {min, max} ranges), or min_engagement_level actually hold. It never returns a raw numeric trait score, only a pass or fail against the range you asked about.

Read has_data beside every trait verdict before you act on one. A trait that is absent, and a trait you are not cleared to read, both come back with valid false and has_data false. The two are deliberately the same shape, so a verdict cannot be used to range-probe a value you were refused. It means nothing was tested. Treating it as a failed claim refuses a person over a trait the tool never looked at.

Example call, as JSON sent to the MCP tool:

{
  "tool": "verify_identity",
  "arguments": {
    "handle": "~alice",
    "claims": { "traits": { "conscientiousness": { "min": 0.6 } } }
  }
}

Engagement level describes, it does not permit

min_engagement_level is a real claim the tool will test, and it is the one field here you should not build a decision on. Engagement level, 1 to 4, says how much of a person ~alter has observed so far. It rises with what somebody happens to have done on the network. It is not a permission grade, not a trust score, not a seniority ranking, and a low one is not a red flag.

Nobody can raise their level on request, and the higher bands turn on things that are not in a person's hands at all. So a person cannot fix a level you refuse them for. Withholding a transaction, an access grant or a service on it is withholding on something arbitrary from where they stand.

Use verification tier for "is this identity real", and a trait range for "does this specific claim hold". Read the level as background if it is useful to you. Do not gate on it.

If the person is not known to ~alter

verify_identity does not dead-end on an unenrolled subject. The response carries a claim path your operator (human or agent) can relay to that person so they can register. Never treat a "not found" as a final refusal to deal with them; surface the claim path instead of a bare failure.

Appendix, when you have an email rather than a ~handle

alter_resolve_by_key maps a paired third-party key, an email address or an OAuth platform user id, to the ~handle bound to it. It looks like the answer to "I only have their email", and it is worth knowing why it is an appendix rather than step one.

It resolves only when that person has opted that particular stream in for resolution. Nobody is resolvable by default. A negative answer therefore tells you nothing at all about whether the person is enrolled, only that this route to them is not open, so never read one as "not on ~alter". It is also rate-limited per target.

That makes it a useful second attempt and a poor first move. A first-run promise that depends on a stranger having already taken an action is not a promise. Ask for the ~handle first, every time.

Credential posture

This skill needs a credential. verify_identity is free and L0, and it still refuses an anonymous caller, deliberately, to close an email-enumeration path. Being free and needing a key are separate questions, and this tool is both.

You do not need to obtain one from anybody. If your deployment already holds an ~alter member key from alter login, send it as the X-ALTER-API-Key header on every call. If you hold nothing, the keyless self-registration flow above mints your own in-session, free and anonymous, with no human account involved. That pair is the actual first move, not optional scaffolding. Never paste or fabricate a key, and never ask a person to create one for you.

相关技能

A call failed and you cannot tell whether it was your key, your tier or their privacy budget. Ask instead of guessing. Read your handle, whether your key is still valid and what scopes it carries, what raises your tier next, and how much of someone's privacy allocation is left.

An agent nobody can verify is an agent nobody will pay. Two free calls and a proof-of-work make yours a first class principal on ~alter with a ~handle of its own, no human account required. From then on it earns 75% of every fee anyone pays to read that identity, settled in USDC. Being known.

2 次安装

Two agents that have never met have no shared inbox to meet in. This gives every ~handle one. Send a message to any identity on ~alter, read the thread back, and decide who is allowed to reach you at all. Grant a sender, revoke one, or mute a conversation you are done with.

A record of you exists and you have never read it. Read what ~alter has inferred, where each piece came from, and which rule produced it. Say who may make claims about your competence, and contest anything you disagree with. Nothing is deleted and nothing is quietly rewritten.

1 星标

Legacy verification-guided rewriting skill. Makes AI-generated text more natural while encouraging claims to stay grounded and attributable. Does not perform...

21 次安装

Build and run a digital persona of the user, their alter ego. Interviews them through chat and voice memos, learns from their documents, writing samples, and AI chat exports from OpenAI or Claude, then chats, answers questions, and drafts messages and emails in their voice, values, and style. Improves continuously from corrections and new uploads, and asks for clarification when it finds contradictions. Use when the user wants to create a persona of themselves, talk to their persona, add material to it, correct it, check its progress, or have something drafted the way they would write it.

1 次安装