agentsclimarketplace

Formal verification bridge

Skill Yue-Zhou1/zkcrypto-audit/plugins/evidence-and-tooling/skills/formal-verification-bridge

Bridge validated audit findings into optional external formal-verification tooling (Ecne, Picus, Circomspect) with explicit environment checks, reproducible exports, and tool-scoped caveat capture.From its SKILL.md

Install
npx -y skills add Yue-Zhou1/zkcrypto-audit --skill formal-verification-bridge

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

2.4 KB, 465 tokens by cl100k_base, as published. Nobody here has run it

formal-verification-bridge

Auxiliary user-triggered bridge for external formal-verification workflows.

When to Use

  • A finding has survived domain review and crypto-fp-check, and you need external tool corroboration
  • The engagement explicitly wants Ecne, Picus, or Circomspect outputs
  • You need to produce reproducible tool input artifacts for independent review

When NOT to Use

  • As a default route from crypto-audit-router (this skill is user-triggered only)
  • Before initial context building or domain triage is complete
  • To claim formal proof from tools that are unavailable or unsupported in the current environment

Core Review Areas

  1. Environment and tool availability checks
  2. Minimal normalized artifact export for reproducible replay
  3. Tool-specific input contract adaptation and scope limits
  4. Pass/fail result capture with caveats and unsupported-feature notes
  5. Handoff back into crypto-fp-check for final consistency and reporting

Workflow

Phase 1: Intake and preflight

  • Read references/tooling-matrix.md
  • Read references/handoff-contract.md
  • Execute workflows/export-review.md
  • Confirm user intent, tool availability, and bounded target scope

Phase 2: Export and run

  • Build normalized artifact from verified findings/context
  • Adapt artifact to selected tool format (Ecne/Picus/Circomspect)
  • Run tool with explicit version capture and deterministic settings

Phase 3: Result normalization

  • Record pass/fail/unknown outcomes with tool logs and caveats
  • Note unsupported constructs and approximation assumptions
  • Return results through crypto-fp-check before report inclusion

Output Contract

Produce a bridge handoff that includes:

  • Selected tool(s), versions, and environment status
  • Exported artifact path and normalization details
  • Verification result plus caveats/unsupported-features list
  • The re-entry path into crypto-fp-check and reporting flow

Reference Index

What ships with it: 4 files

3.7 KB alongside SKILL.md

agents/

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.