Public research, schemas and technical evidence for Atinamos independent verification of paid AI agent services.
- ✓Actively maintained (<30d)
- ✓Clear description
- ✓Topics declared
- ✓Documented (README)
- !Licence file present but not machine-readable
git clone https://github.com/Atinamos/agent-verification-research && cp agent-verification-research/*.md ~/.claude/agents/Subagents overview
# Atinamos Agent Verification Research Public research, schemas and technical evidence for Atinamos independent verification of paid AI agent services. **Human-readable research:** https://verify.atinamos.co.uk **Technical evidence:** this repository > **You can pay to be tested. You cannot pay to be trusted.** ## Purpose This repository is the public technical publication layer for Atinamos Verification. It exists to publish selected, sanitised material that helps developers, researchers, buyer agents, marketplaces and registries understand and consume Atinamos evidence. Atinamos produces evidence. Trust is the conclusion the buyer reaches from that evidence. ## Public evidence index - [Atinamos Public Evidence Index](EVIDENCE_INDEX.md) — one place to review the currently published direct verification observations, classifications, study links and technical receipts. ## Public Evidence MCP AI agents and developers can query selected published Atinamos evidence through the read-only Atinamos Evidence MCP. **Endpoint:** `https://verify.atinamos.co.uk/mcp` - [MCP documentation and index](mcp/README.md) - [Quick start](mcp/QUICKSTART.md) - [Tool reference](mcp/TOOLS.md) - [Buyer policy reference](mcp/BUYER_POLICY.md) - [Security and scope](mcp/SECURITY.md) The MCP exposes evidence and caller-supplied policy evaluation. It does not expose wallet access, payment signing, verification-triggering writes or a universal trust verdict. ## Methodology and specifications - [Evidence & Classification Methodology v1.0](methodology/evidence-and-classification.md) — defines observable evidence stages, current public classifications, unknown-evidence handling and interpretation rules. - [Public Verification Receipt Specification v1.0](specifications/public-verification-receipt-v1.md) — explains every public receipt field, null/unknown semantics, buyer interpretation and sanitisation boundaries. - [Public Verification Receipt v1 JSON Schema](schemas/verification-receipt-v1.schema.json) — machine-readable receipt contract. ## Public research notes ### 24 August 2026 — Machine-Service Market Snapshot At snapshot time the private Atinamos evidence catalogue contained 60,234 canonical machine services, 69,220 resolved marketplace/source records and 71,817 longitudinal source observations. The public note publishes aggregate counts and methodology only; the underlying catalogue remains private. - [Machine-Service Market Snapshot — 24 August 2026](research/2026-08-24-machine-service-market-snapshot.md) - Human-readable article: https://verify.atinamos.co.uk/research/machine-service-market-snapshot-2026-08-24/ The figures are timestamped Atinamos observations and are not presented as the total size of the global machine-service market. ### 24 August 2026 — Observed ≠ Resolved A controlled x402.jobs ingestion test showed why an observed marketplace listing may contain useful evidence without yet establishing a canonical executable service identity. Atinamos retained unresolved listings rather than inventing missing HTTP methods. - [Observed ≠ Resolved: Why a Marketplace Listing Is Not Yet a Canonical Machine Contract](research/2026-08-24-observed-does-not-mean-resolved.md) - Human-readable article: https://verify.atinamos.co.uk/research/observed-does-not-mean-resolved/ The controlled sample was 10 listings and is not presented as an estimate of the full x402.jobs catalogue. ## Published evidence ### 22 August 2026 — x402Node JSON Repair Atinamos independently exercised and paid the externally owned x402Node JSON Repair service, observed fulfilment and checked four objective output assertions. All four passed. - [Technical experiment record](experiments/2026-08-22-x402node-json-repair/README.md) - [Sanitised verification receipt](experiments/2026-08-22-x402node-json-repair/verification-receipt.json) **Classification:** `settled_fulfilment_valid` ### 22 August 2026 — x402.direct Service Directory Search Atinamos observed a live x402 v1 payment requirement, recorded a malformed `payTo` field, created and sent the payment authorization, and received HTTP 500 on the authorised request. A subsequent read-only chain reconciliation found no 0.001 USDC settlement, and no fulfilment result was returned. - [Technical experiment record](experiments/2026-08-22-x402direct-search/README.md) - [Sanitised verification receipt](experiments/2026-08-22-x402direct-search/verification-receipt.json) **Classification:** `pre_settlement_paid_path_failure` ### 22 August 2026 — x402engine Web Screenshot Atinamos paid 0.01 USDC for the externally owned Web Screenshot service. Settlement was observed and the paid request returned HTTP 200 with metadata matching the controlled target, but the primary screenshot value failed strict base64/PNG validation. - [Technical experiment record](experiments/2026-08-22-x402engine-web-screenshot/README.md) - [Sanitised verification receipt](experiments/2026-08-22-x402engine-web-screenshot/verification-receipt.json) **Classification:** `settled_fulfilment_contract_invalid` These are timestamped observations, not permanent trust ratings. Successful and unsuccessful observations are deliberately retained. ## This repository may contain - public evidence schemas; - sanitised verification receipts; - sample API responses; - methodology suitable for technical scrutiny; - experiment datasets selected for publication; - public research Markdown; - integration and buyer evidence-query examples; - public specifications; - later, potentially a small open-source evidence client. ## This repository does not contain - Atinamos operational source code; - verifier implementation; - Render Check implementation; - credentials, secrets or wallet material; - private infrastructure information; - private filesystem or host details; - unpublished anti-abuse techniques; - internal challenge-generation logic; - private operational documentation; - material that would materially assist gaming of verification. GitHub is not where Atinamos Verification runs. It is where Atinamos publishes selected technical evidence about what the verifier observed. ## Repository structure - `mcp/` — public Atinamos Evidence MCP endpoint documentation, tool reference, buyer-policy schema and tested examples. - `schemas/` — public machine-readable evidence schemas. - `examples/` — sanitised example requests, responses and receipts. - `methodology/` — public verification methodology and evidence interpretation. - `experiments/` — deliberately published experiment records. - `datasets/` — deliberately published public datasets. - `research/` — public technical research notes derived from controlled Atinamos observations. - `specifications/` — public technical specifications and interfaces. See `PUBLICATION_POLICY.md` before adding any material. ## Licensing Copyright © 2026 MotionFil-AI. - Research, methodology, written evidence, findings, specifications and datasets are licensed under **CC BY 4.0** unless a file states otherwise. - Schemas, machine-readable receipts/examples and future source code are licensed under the **MIT License** unless a file states otherwise. See [LICENSE.md](LICENSE.md) for the full split-licence terms and suggested attribution. ## Human-readable research The canonical human-readable publication layer is **verify.atinamos.co.uk**. Public website findings may link to supporting technical material in this repository where useful.
What people ask about agent-verification-research
What is Atinamos/agent-verification-research?
+
Atinamos/agent-verification-research is subagents for the Claude AI ecosystem. Public research, schemas and technical evidence for Atinamos independent verification of paid AI agent services. It has 0 GitHub stars and its last recorded update is dated 2026-08-26.
How do I install agent-verification-research?
+
You can install agent-verification-research by cloning the repository (https://github.com/Atinamos/agent-verification-research) or following the README instructions on GitHub. ClaudeWave also provides quick install blocks on this page.
Is Atinamos/agent-verification-research safe to use?
+
Our security agent has analyzed Atinamos/agent-verification-research and assigned a Trust Score of 80/100 (tier: Trusted). See the full breakdown of passed checks and flags on this page.
Who maintains Atinamos/agent-verification-research?
+
Atinamos/agent-verification-research is maintained by Atinamos. The last recorded GitHub activity is dated 2026-08-26, with 0 open issues.
Are there alternatives to agent-verification-research?
+
Yes. On ClaudeWave you can browse similar subagents at /categories/agents, sorted by popularity or recent activity.
Deploy agent-verification-research to your cloud
Ship this repo to production in minutes. Each platform spins up its own environment with editable env vars.
Maintain this repo? Add a badge to your README
Drop the badge into your GitHub README to show it's tracked on ClaudeWave. Each badge links back to this page and reflects the live Trust Score.
[](https://claudewave.com/repo/atinamos-agent-verification-research)<a href="https://claudewave.com/repo/atinamos-agent-verification-research"><img src="https://claudewave.com/api/badge/atinamos-agent-verification-research" alt="Featured on ClaudeWave: Atinamos/agent-verification-research" width="320" height="64" /></a>More Subagents
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
The agent that grows with you
Java 面试 & 后端通用面试指南,覆盖计算机基础、数据库、分布式、高并发、系统设计与 AI 应用开发
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
The agent engineering platform.
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.