Skip to content

den0th/aisthetic-hard-test

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

aisthetic-hard-test

Blackbox stress-test agent for aisthetic.services — a verifiable commerce control plane for AI agents on Solana.

This 200-line agent hammers the public surfaces of the platform with 26 scenarios across four acts. Every successful action produces a cryptographically signed receipt. Every receipt is independently verifiable.

                                                            ┌─────────────────┐
hard-test agent ─── 26 scenarios ──┬──→ sandbox.aisthetic ──┤ 7-stage pipeline│ ──→ signed receipts
                                   │                        └─────────────────┘
                                   └──→ Jupiter v6 ─────────→ Solana mainnet  ──→ finalized tx

What it does

Act Scenarios What it tests
Coverage 15 liveness probes, 401 → 402 → 200 paid-call chain, receipt id shape, receipt verifier page, activity feed propagation, all public landing routes (/explorer, /docs/api, /proof-center, /agent/[hash], /principal/[hash])
Adversarial 10 replay/dedupe, malformed Authorization, wrong sandbox-proof never-200 invariant, oversized URL, path traversal, SQL pattern in bearer, random/empty bearer, unknown gateway provider, PII guard on 50-row activity feed
pay.sh interop 4 probe Solana Foundation's pay.sh debugger 402 challenge, parse WWW-Authenticate: Payment header, decode the base64 MPP challenge body, assert that aisthetic and pay.sh both expose machine-readable Solana payment requirements at the 402 layer
Stress 1 200 concurrent paid calls — uniqueness invariant (every successful receipt MUST be unique under load), p50/p95/p99 latency capture
Trade 4 pre-trade gateway receipt, real Jupiter quote on mainnet, signed Solana swap (0.001 SOL → USDC, ~$0.09) finalized on-chain, post-trade gateway receipt

Latest run

See reports/ for the most recent run output. Verdict: 30/30 PASS, 64 receipts produced, real mainnet Solana tx finalized.

Each receipt is a real signed claim — verify any one of them at https://aisthetic.services/r?id=<receipt-id>.

Run it yourself

Prereqs

  • Node 20+ and pnpm
  • (Optional, for Act 4) A Solana mainnet keypair with ≥ 0.01 SOL at ~/.at-canary/agenttrust-mainnet.json

Install

pnpm install

Run

# All 4 acts including real mainnet trade (~$0.10 in SOL)
pnpm test

# Skip the trade act (no real SOL spent)
pnpm test:no-trade

The report writes to reports/<date>-hard-test.md.

Why this exists

Trust is the bottleneck for autonomous AI agents transacting online. aisthetic.services is the verification layer — every agent move produces a receipt that anyone can check, every payment settles in stablecoin without a custodian, every agent gets a portable on-chain identity.

This repo is the public proof. Don't trust us — run the agent. It uses only public surfaces, only documented headers, only signed payloads.

What this repo doesn't include

  • The platform source code (private)
  • Any operator credentials, API keys, or signing keys
  • Anything that isn't already a public HTTP surface

Ownership

Don't trust this repo because it claims to be aisthetic. Verify the signer:

Related

License

MIT. See LICENSE.

About

Blackbox stress-test agent for aisthetic.services — 26 scenarios across 4 acts, real Solana mainnet trade, every receipt independently verifiable.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors