agentsclimarketplace
Sign in

Fault injection auditor

Skill Yue-Zhou1/zkcrypto-audit/plugins/implementation-safety/skills/fault-injection-auditor

Audit cryptographic code for active fault-injection attacks: RSA-CRT Bellcore faults, ECDSA/EdDSA differential fault analysis, verification-skip glitches, redundant-computation bypass, and verify-after-sign gaps. Use when the threat model includes an attacker who can glitch computation, distinct from passive side-channel leakage.From its SKILL.md

Install
npx -y skills add Yue-Zhou1/zkcrypto-audit --skill fault-injection-auditor

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.8 KB, 785 tokens by cl100k_base, as published. Nobody here has run it

fault-injection-auditor

Domain auditor for ACTIVE fault attacks — an adversary who can corrupt a computation (voltage/clock glitch, laser, rowhammer) and observe the faulty output. This is a separate discipline from passive side channels: the attacker changes the computation rather than only measuring it.

Dedicated skill by design: fault resistance is not folded into side-channel-auditor's timing/cache/power scope.

When to Use

  • The threat model includes physical or software-induced fault injection
  • Auditing RSA-CRT signing for Bellcore faults
  • Auditing deterministic ECDSA/EdDSA signers for differential fault analysis
  • Reviewing verification-skip glitches, redundant-computation bypass, and verify-after-sign presence
  • Assessing fault-detection countermeasures and their coverage

When NOT to Use

  • Passive timing/cache/power leakage (attacker only measures) -> side-channel-auditor
  • Signature verification-equation or encoding correctness -> signature-scheme-auditor
  • Randomness/nonce lifecycle -> randomness-auditor
  • Pure protocol-correctness review with no physical/fault attacker

Core Review Areas

  1. Fault model: what the assumed attacker can do (single/multiple faults, instruction skip, data corruption, targeting precision) — findings are only meaningful relative to a stated model
  2. RSA-CRT: verify-after-sign OR redundant recombination; a single fault in one CRT half factors N (Bellcore attack)
  3. Deterministic signatures: DFA on ECDSA/EdDSA — a correct+faulted pair over the same message recovers the key; countermeasure is redundancy or re-verification
  4. Verification skip: can a glitch bypass the branch that enforces a signature/MAC/proof check (skip the if (!valid) reject)?
  5. Redundant computation and consistency checks: presence, coverage, and whether the check itself is fault-attackable (double-fault)
  6. Verify-after-sign: signer re-verifies its own output before release
  7. Evidence limits: fault findings are usually theoretical without lab hardware — disposition honesty (observation/residual_risk) matters

Workflow

Phase 1: Fault-model and surface mapping

  • Read references/fault-injection-checklist.md
  • Record the assumed attacker model and enumerate fault-sensitive operations (CRT, deterministic signing, verification branches)

Phase 2: Fault-path review

  • Execute workflows/fault-path-review.md

Phase 3: Pattern hunt

  • Read references/finding-patterns.md
  • Prioritize missing verify-after-sign, unprotected CRT, and single-branch verification

Phase 4: Handoff

  • Send surviving findings to crypto-fp-check with explicit evidence limitations
  • Cross-route passive leakage to side-channel-auditor and signature semantics to signature-scheme-auditor

Output Contract

Produce a fault-injection handoff that includes:

  • fault_model
  • injection_point
  • redundancy_or_detection_invariant
  • evidence_limitations
  • disposition (one of verified, false_positive, unverified, observation, residual_risk)
  • next_route

Reference Index

What ships with it: 5 files

10.2 KB alongside SKILL.md

agents/

references/

workflows/

Keep looking

Skills are one crate of 326,871. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.