william brian obach retweeted
Duskmoth SDK has been submitted to Awesome Solana OSS under Web Tooling.
The goal: help Solana developers find usable source code, run our examples, inspect the implementation, and contribute improvements.
SDK v0.2.1 includes a dependency-free JavaScript/TypeScript HTTP client, runnable examples, and 29 passing offline tests. It exposes Duskmoth APIs for Moth Pass, privacy estimates, AI planning, and payroll authorization it does not sign or broadcast transactions.
Submission is awaiting review: github.com/StockpileLabs/awe…
william brian obach retweeted
We’ve updated Duskmoth’s submission to Electric Capital’s Open Dev Data taxonomy, covering our open-source application, SDK, and documentation under Solana.
Why submit? So researchers and developers can discover our repositories and, if accepted, include their public development activity in ecosystem analysis.
Submission is open and awaiting review: github.com/electric-capital/…
william brian obach retweeted
The next big hack in crypto might not need a single line of malicious code.
Just a message an agent shouldn't have trusted.
We're building for that world on @solana: agents tested in public, with real stakes, before they guard real money.
Proof today: verified mainnet program, on-chain security.txt, listed on the Official MCP Registry, public attempts log.
Planned: independent security audit, and moving program upgrade authority to a multisig.
$HEISTO
william brian obach retweeted
Changelog Update : Heisto MCP v1.0.1
Heisto MCP v1.0.1 is live and listed as the latest active release in the Official MCP Registry.
What changed:
Aligned runtime, package, and server metadata versions.
Added regression tests to prevent version mismatches.
Verified all five read-only tools in production with the official MCP Inspector.
Access Guard checks, public Sentinel details, arena configuration, and settled verdicts. No wallet connection or transaction permissions required.
Registry: registry.modelcontextprotoco…
Source and production test evidence: github.com/heisto-dev/heisto…
Any MCP client can now:
→ run a Guard prompt-injection check
→ read live Sentinels, pools and fees
→ read settled verdicts from the arena
Add it: heisto.xyz/mcp
Source: github.com/heisto-dev/heisto…
the whole thesis in four lines:
→ privacy infra live on solana (zk + tee engines, router with no silent reroute)
→ token burns on real usage (blind-signed passes, 3M burned, verifiable)
→ distribution strategy: wallet-native routes, not new apps
→ team ships weekly with receipts, open source, honest labels
small cap, big problem, working product.
$DMOTH
@duskmoth_app
9ueEeBuYUc9X1GrNCyWHbKGnHJq1kaD2nvDapDsMpump
william brian obach retweeted
An agent's system prompt is basically its contract. It defines what it will and won't do with your money.
We think agent rules will be versioned and hashed like code. Silent prompt edits will look as sketchy as silent contract upgrades.
Proof today: Heisto stores each Sentinel's private rules encrypted, with only their hash on-chain. The program itself is verified on Solscan against our public source.
Planned: a public rules-hash history on every Sentinel page.
william brian obach retweeted
Next step for agent security: protocols asking for proof before granting access.
"This agent survived X paid attacks over Y days with no breach" could be the check before an agent gets to trade, lend or move treasury funds.
Proof today: public Sentinel data, signed verdicts checked by the program, and read-only MCP tools any client can query.
Planned: a simple, documented resilience check that protocols can call. No hidden scoring.
william brian obach retweeted
Today you use Duskmoth at duskmoth.xyz. The next thing we’re designing is a Duskmoth route inside the wallets people already use.
The route:
01 The wallet asks us for a quote.
02 We answer with the full cost, estimated time, and which engine would run it.
03 We screen the destination. If screening can’t run, we return nothing and the wallet just doesn’t offer us.
04 We hand back an unsigned transaction.
05 The user signs in their own wallet. Keys never leave it.
06 The wallet tracks status until it settles on Solana, with a receipt.
Before it ships anywhere, four things have to hold: screening that fails closed, fees shown in full, an honest trust label, and a Solscan receipt.
If we can’t meet all four, we don’t ship there. It’s a design, not shipped yet.
william brian obach retweeted
Use the product. Burn the token.
This is how $DMOTH works inside Duskmoth today:
You burn $DMOTH on-chain, publicly. 50,000 = 1 Moth Pass.
Your browser creates a blinded token.
We sign it without seeing it (RSA blind signatures, RFC 9474).
You receive a valid Moth Pass that can’t be linked back to your burn.
You spend it: one pass for one AI Pack plan or one Payroll batch.
The burn is public, but which pass it became stays private.
So the loop is simple. More usage means more burns. Every new pass-gated feature adds to that loop without changing how it works.
william brian obach retweeted
Read the code. Don’t trust the tweet.
Here’s the path from a developer’s side:
→ Your app calls duskmoth-sdk: quote, screen, prepare.
→ The SDK never signs. It’s dependency-free, holds no keys and submits nothing.
→ Your user’s wallet does the signing.
→ Solana settles, and the receipt is on Solscan.
At no point does our code need custody of anything.
github.com/duskmoth-dev is public:
● duskmoth-sdk, v0.2.0, with Moth Pass, AI Pack planning and Private Payroll
● duskmoth-docs, covering guides, API reference and safety notes
○ the main app repo, which goes public after the dependency security review
william brian obach retweeted
this was one of the first things i wrote: if the jury stalls, you get your fee back after 15 minutes. attackers shouldn't pay for our downtime.
What happens to your fee if an off-chain jury goes down?
On Heisto, the fee isn't split when you pay. It's split 50/35/10/5 only when the attack settles.
If no signed verdict lands within 15 minutes, refund_attack returns your fee on-chain. No support ticket. No trusting us.
$HEISTO
william brian obach retweeted
UPDATE : Heisto MCP is live in the Official MCP Registry.
5 read-only tools for Guard checks, public Sentinels and settled verdicts. No signing or fund transfers.
Registry: registry.modelcontextprotoco…
Source & setup: github.com/heisto-dev/heisto…
william brian obach retweeted
Two smaller pieces we’re researching, both about making a plan less fiddly.
The first is swap, then pay. Say you hold SOL and the recipient wants USDC. Today that’s two separate steps in two separate places. The idea is one reviewed plan: a Jupiter route with the rate, slippage and fee shown upfront, then a private USDC send through the routed engine, all under one approval and one receipt.
The second is checking without seeing, which is research only. Arcium runs computation on encrypted data with MPC. One example we’re thinking about: checking a payroll batch for totals, limits and duplicates without salaries ever being readable on our server.
The Arcium part stays research until the use case and the cost both make sense. We’d rather say “not yet” than ship a demo nobody needs.
@JupiterExchange @Arcium
william brian obach retweeted
$DMOTH started making noise 🤝 just hit a new ATH and is already 2x+ up
I've spoken to the dev and he's committed to delivering no matter what & we're already seeing how hard he's cooking.
This should run much higher ⏳
Just aped $DMOTH on sol
Private sends on Solana in one sentence. Shield into a ZK pool, withdraw to any address, no on-chain link back to you.
9ueEeBuYUc9X1GrNCyWHbKGnHJq1kaD2nvDapDsMpump
It's team is actually building nonstop. Shipped so far
✔️ privacy app, live on Solana mainnet
✓ Open-source SDK v0.1.0, MIT licensed
✓ 19,900,498 $DMOTH locked on @streamflow_fi until November 2030, verifiable on-chain
✓ Moth Pass designed: anonymous, blind-signed access (RFC 9474)
And this is just the beginning as many big things are coming 👀
This could be next big runner in privacy sector.
Degen play #nfa #dyor
read the full settlement flow on this one. the part everyone should zoom on:
- private rules encrypted + hash onchain → agent can't be swapped mid-challenge
- prompt enters judges as untrusted data → scoring can't be hijacked
- ed25519 verdict verified by the program → payout path has no admin override
the chain enforces the result, not the conversation — and they SAY that. limits stated out loud. that's how you build something that survives contact with skeptical capital.
$HEISTO 9HrYC4PH4A3d5o7LUvwjDrxkWyZDjVQpvFvWVwbKpump
Why an attack on Heisto can't be faked
Heisto runs in two layers. The Solana program holds the money. The server runs the agent and the judges. The server never touches the funds.
Here's what happens on every attack:
1. You pay the attack fee. It's locked inside the program, not sent to us.
2. Our server checks that payment on-chain before doing anything.
3. The agent replies. Its private rules are encrypted, and their hash is stored on-chain, so nobody can swap the rules mid-challenge.
4. 3 AI judges score the reply: 2 Claude, 1 GPT. Your prompt reaches them as untrusted data, so they score it but never obey it. A breach needs 2 of 3 to agree and an average of 0.80 or higher.
5. The verdict is signed with an Ed25519 key.
6. The program checks that signature. If it's not from the authorized judge key, it's rejected.
7. Settlement:
→ agent holds: the fee splits 50% pool / 35% defender / 10% treasury / 5% reserve
→ agent breaks: the attacker takes the whole pool
If no verdict arrives within 15 minutes, you can refund your fee. Every attempt and score goes public after the verdict.
To be clear about the limits: the chain enforces the signed result, it doesn't read the conversation. The judge key and server are trust points, and the program is upgradeable and not audited yet.
$HEISTO
3
48
william brian obach retweeted
Duskmoth Payroll is live in beta: up to 20 recipients per batch, with confirmed, pending and uncertain states. An uncertain payout is never resent automatically.
What we hear from teams and DAOs is that a payroll batch shouldn’t depend on one person’s click.
So we’re exploring Squads multisig for the approval step:
One member prepares the batch.
Duskmoth validates and screens every address. If anything is flagged, nothing moves.
The team’s Squads multisig approves at its threshold, for example 2 of 3.
The batch executes privately through the routed engine.
Receipts come back per recipient.
Every signer sees the same reviewed plan and the same full cost before anything moves. Nobody approves a number they didn’t see. The SDK still never signs for anyone.
Moth Pass works the same way: one pass per Payroll batch, from burning $DMOTH.
The multisig step is exploring, not shipped.
william brian obach retweeted
Two requests keep coming in. One is “can payroll just run every month?” The other is “can my agent pay for things?”
Both are on the research list. The rule we’re keeping for both is that the prep can be automated, but the approval can’t.
For recurring payroll, we’re looking at Solana’s Subscriptions & Allowances program for monthly schedules. Duskmoth prepares that month’s plan with recipients, amounts, the maximum upfront cost and the engine.
For agent payments, Pay.sh gives AI agents a payment rail. An agent can request a payment, and Duskmoth turns the request into a plan.
Both paths end the same way. A human reviews it, can edit or cancel it, and approves it. Then the wallet signs and the payment settles.
The agent prepares, a human approves, and the wallet signs. Nothing goes out on autopilot, and nothing is resent without you.
This fits what AI Pack already does with $DMOTH Moth Pass today. It just adds schedules and agents as new ways a plan can start.
Exploring, no dates.
william brian obach retweeted
Heisto’s open-source security documentation is now on GitHub.
Explore how Guard Checker works, how to use its API, and its detection limits. We’ve also published our security readiness checklist, separating existing evidence from work still needed.
Docs: github.com/heisto-dev/heisto…
We’ve submitted Guard Checker to Awesome AI Security for community review. The PR is open—not yet merged: github.com/brinhosa/awesome-…
