Openclaw cli: OpenClaw reuses the gateway auth token in the owner ID prompt hashing fallback (CVE-2026-32897)

The OpenClaw CLI vulnerability (CVE-2026-32897) involves reuse of the gateway authentication token (`gateway.auth.token` and `gateway.remote.token`) as a fallback secret for hashing the owner ID prompt when `commands.ownerDisplay=hash` and `commands.ownerDisplaySecret` is unset. This creates a secret dual-use scenario between gateway authentication and prompt metadata hashing. Hash outputs are visible to third-party model providers in system prompts, but no direct plaintext token disclosure occurs. The risk is higher if operators use weak gateway tokens and do not set a dedicated owner hash secret. The issue affects OpenClaw CLI versions up to 2026.2.21-2. The planned fix in version 2026.2.22 removes the fallback to the gateway tokens and introduces auto-generation and persistence of a dedicated `commands.ownerDisplaySecret` when hash mode is enabled and no secret is set.

The vulnerability causes the gateway authentication token to be reused in a different security context, increasing the risk of token compromise if hash outputs are observed by third parties. Although the token is not directly disclosed in plaintext, the dual-use of the secret across authentication and prompt metadata hashing domains weakens security boundaries. The practical risk is highest when weak gateway tokens are used and the owner hash secret is not configured, potentially allowing attackers with access to prompt metadata to infer authentication secrets.

A patch is available and planned for release in OpenClaw CLI version 2026.2.22. This patch removes the fallback use of the gateway authentication tokens for owner ID prompt hashing and introduces a dedicated secret that is auto-generated and persisted when hash mode is enabled and no secret is set. Operators should upgrade to version 2026.2.22 once released to remediate this issue. Until then, setting a strong `commands.ownerDisplaySecret` can reduce risk.