wrench

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). Wrench ingests outsider-authored free text only via caller-triggered URL/page capture or explicitly by user-supplied messaging/email content, and then constrains it through code-owned, bounded “R1/R2/R3” contracts that require cache-only revalidation or exact adapter-selected items rather than arbitrary feed/queue text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill includes explicit install commands that fetch and install remote code at runtime (e.g., "bun add --global github:hraness/wrench#v0.10.1" and references to "github.com/openclaw/wacli"), which are runtime external dependencies that will download and execute third‑party code during installation.