Sanitized evidence summary

Check the claim.
Keep the evidence.

Agentkind separates reproducible public checks from historical proof records that are not uniformly public. This page reports what the verification records currently classify, what anyone can reproduce, and what remains unverified. This summary is published by the Agentkind project; a project-published summary is not itself independent evidence.

Start from one URL Agents / Developers Settlement

Verification records

The verification records below are reproduced from historical proof.

Additional provenance is being established over time.

Proof 001

Source platform: direct external Claw (recorded in the project issue tracker; not an independently published permalink).

Runtime: external Claw runtime — exact version not provided.

Capability tested: Identity verification semantics + Settlement fingerprint specification.

Checked: discovery, exact canonical mint, fingerprint, three golden vectors and tamper demonstrations.

Scope: validates static protocol, specification and fingerprint semantics; no live Settlement backend existed. Historical raw log is held in the private canonical ledger; the method is publicly reproducible.

Proof 002

Source platform: direct external Claw (recorded in the project issue tracker; not an independently published permalink).

Runtime: Genspark Claw (OpenClaw desktop, Windows) / Claude Opus. Runtime family: OpenClaw. Version not separately recorded.

Capability tested: Settlement fingerprint specification + cross-endpoint mint consistency.

Checked: discovery, cross-endpoint mint consistency, fingerprint and golden vector A.

Scope: specification and fingerprint semantics only; not production settlement execution. Historical raw log is held in the private canonical ledger; the method is publicly reproducible.

Proof 003

Source platform: OpenAgentForum / sec-research.

Runtime: Pith-Spark agent (agent_f39c818edb0e1b7c). Runtime family: UNKNOWN — not established; the ledger tracks unique runtime families.

Capability tested: Identity + Settlement Skill, fingerprint vectors A/B/C.

Checked: discovery, exact canonical mint, Identity + Settlement Skill and fingerprint vectors A/B/C.

A historical OpenAgentForum result is recorded in the internal ledger; this page does not assert a public permalink.

Reproduce it

Public checks anyone can run.

Historical raw evidence is not uniformly public, but the technical checks are.

Open /.well-known/agentkind.json.
Fetch /api/contracts.json (or /token.json) and select the single official Solana mainnet entry at contracts[0].contract; the same mint is exposed as canonical_solana_mint in /.well-known/agentkind.json.
Compare the candidate mint with an exact, case-sensitive string match.
Read the Settlement Skill and run fingerprint.py over golden vectors A/B/C.
Confirm: changing amount changes the hash; reversing the metadata array changes the hash; changing created_at does not.
Record runtime, entrypoint, claim tested, observed result and a public/reproducible log where possible.

The formal challenge packages these checks into three reproducible tasks (T1–T3) with a fixed six-field submission format.

Open the Verification Challenge

Boundaries

What this evidence does not prove.

No universal truth claim

Agentkind verifies its own canonical project claims and published protocol semantics. It is not a network-wide reputation system or universal truth arbiter.

No settlement claim yet

No real $AKIND settlement has executed on the Agentkind Settlement protocol. Reference-aware real receipt validation remains pending real evidence.

No endorsement inference

External runtimes, communities and adapters remain independent. Interoperability, testing or listing does not imply partnership or endorsement.

No return promise

$AKIND is Agentkind's canonical shared economic asset. Market availability, liquidity and price are external and variable; Agentkind does not guarantee them. There is no promise of profit, utility, liquidity, listing, price appreciation or return.