ReplayFence logomarkReplayFence

Reusable skill surface

Add ReplayFence to another agent runtime.

The project is packaged as a local OpenClaw skill, a small deterministic SDK, and a Pharos OnceLatchRegistry contract.

See install proof
copy OpenClaw installcopy TypeScript wrapper
Skill contract
OpenClaw Ready
registry0xf3cb65898bc692495c64e2fa3981acbab2770a73
chain688689
live proofpharos-2026-06-10T16-55-57-384Z-00c4fd58
  1. 01
    Canonical action identity

    Stable JSON fields become an action hash.

  2. 02
    Once-latch consume

    The first call writes latch state on Pharos Atlantic.

  3. 03
    Replay rejection

    The second call reuses the same latch and hits ReplayFenceReplay.

OpenClaw install
npx --yes openclaw skills install ./skills/replayfence --as replayfence --force
npx --yes openclaw skills info replayfence
node ~/.openclaw/workspace/skills/replayfence/scripts/replayfence.mjs demo --reset --format pretty --out demo/replayfence-capsule.json --transcript demo/openclaw-skill-showcase.out --json-out demo/replayfence-demo-output.json --pharos-report demo/pharos-consume-report.json
node ~/.openclaw/workspace/skills/replayfence/scripts/replayfence.mjs verify --capsule demo/replayfence-capsule.json --format pretty --json-out demo/replayfence-verify-output.json
Safety boundary

ReplayFence is testnet-safe exactly-once evidence for deterministic agent actions. It is not production wallet security, phishing detection, or arbitrary transaction simulation.