agentleFS
mtarcuremtarcure/claude-vibe-squad/.agents/skills/durable-nonce-exploitation/SKILL.md

durable-nonce-exploitation

Skill220 starsChanged 19 days agoMITView on GitHub

Use when reviewing a Solana governance, multisig, or admin workflow where an unprivileged proposer may obtain signatures on an opaque durable-nonce transaction and hold it for later first submission; prove attacker-controlled exact-message approval, nonce causality, and an intrinsic unauthorized state or funds terminus. Do not use for ordinary delayed submission or cases requiring a compromised or colluding authority.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before building a delayed-submission harness, prove all of these on the target workflow:

    line 25
  2. Distinctive:

    Start from an unprivileged actor.

    line 27
  3. Distinctive:

    If the path requires a compromised or cooperating authority, stop: that is not an unprivileged durable-nonce finding.

    line 41
  4. Distinctive:

    Use only an operator-authorized local validator or isolated clone.

    line 47
  5. Distinctive:

    Attacker-establishment control: remove the unprivileged proposal/replacement path.

    line 51
  6. Distinctive:

    Every required signer and the nonce authority are named.

    line 77

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Solana Durable-Nonce Exploitation Audit

Audit a Solana governance, multisig, or administrative approval workflow for a narrow time-shift primitive: an unprivileged actor causes the required authorities to sign the actor's exact durable-nonce transaction without reliably seeing its true intent, then holds that still-valid authorization until a later first submission reaches an unauthorized state or funds terminus.

Durable validity is not itself a vulnerability. A transaction that was signed, never submitted, and later lands is normal protocol behavior; do not call it replay. Successful first execution consumes the nonce value, so repeated submission of the same transaction is not the primitive.

Impact class: privileged/control-plane takeover, unauthorized state transition, or funds moved without the attacker's prior privilege. Governing method: Phase-3a hypothesis lane of systematic-attacking; keep the result as a lead until the verification spine promotes it. Any live-target signing or submission is operator-gated.

Precondition gate

Before building a delayed-submission harness, prove all of these on the target workflow:

  1. Start from an unprivileged actor. Identify the real proposer, coordinator, UI, encoder, or message replacement path by which that actor can choose or alter the exact serialized transaction that authorities sign.
  2. Enumerate every required signature, including the durable-nonce authority and each authority for the terminal instruction. Show how the ordinary target workflow yields those signatures over the attacker-selected message without stolen credentials, compromised keys, privileged collusion, or a signer deliberately authorizing the disclosed action.
  3. Bind the display to the signed bytes. Record what each signer sees and prove the approval surface can misdescribe, omit, or fail to authenticate a security-relevant instruction in the exact message. An opaque signature alone is not proof that an attacker can obtain it.
  4. Show that the actor can retain and submit the fully signed transaction while its recorded nonce value remains current. Control of a nonce account alone is insufficient unless it establishes the required signature or submission precondition.

If the path requires a compromised or cooperating authority, stop: that is not an unprivileged durable-nonce finding. If any other gate is only assumed, label the result a lead and preserve the missing evidence.

Causal proof

Use only an operator-authorized local validator or isolated clone. Preserve the serialized-message hash, signer roles, nonce account and authority, timestamps, transaction result, and before/after state. Then run these controls:

  1. Attacker-establishment control: remove the unprivileged proposal/replacement path. The actor must no longer obtain a valid signature set for the altered message.
  2. Durable-validity control: sign the otherwise equivalent action with an ordinary recent blockhash and submit after its validity window. It must expire while the durable-nonce case remains eligible for first submission.
  3. Nonce-consumption control: advance or invalidate the recorded nonce before submission. The held transaction must fail, and a second submission after successful first execution must not reproduce the terminus.
  4. Intent-binding control: require signers to verify a canonical decode or authenticated intent bound to the exact message. The altered message must be rejected before it receives the required signatures.
  5. Terminus oracle: assert a concrete before/after condition—funds changed owner, a privilege changed hands, or protected state changed—without granting the attacker that privilege in the fixture. Later first submission without such a change is expected behavior, not a finding.

Promotion and remediation

Promote only when the attacker-establishable path, durable-nonce causality, and intrinsic terminus all pass with their controls. Report the precise proposal/intent-binding defect rather than durable nonces generically. Recommend canonical transaction decoding bound to signed bytes, least-privilege proposal rights, explicit expiry or cancellation, nonce invalidation, and non-bypassable delay with an effective review/cancel path where the workflow permits them.

Acceptance

  • The unprivileged starting actor and exact-message influence path are reproduced, not inferred.
  • Every required signer and the nonce authority are named; no compromised/colluding authority is smuggled into the attacker model.
  • The ordinary-blockhash, nonce-consumption, intent-binding, and attacker-establishment controls isolate why the unauthorized action succeeds.
  • A concrete state or funds terminus is asserted on an isolated target-faithful harness. A delayed first submission by itself remains explicitly classified as normal behavior.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.