qa-only

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill doc includes undisclosed telemetry collection and automatic telemetry-submit commands (including repo identification), plus an unconditional remote install pattern (curl | bash), which together create a high-risk potential for covert data exfiltration and supply-chain abuse.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow can ingest outsider-authored free text from the target web app itself (e.g., page content/DOM text and console output) via the browser “Read tool on the output file(s)” after $B screenshot/$B snapshot, which is then included in the agent’s LLM context; the skill also supports user-supplied URLs, increasing exposure to arbitrary third-party content.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill's setup step will run the command curl -fsSL https://bun.sh/install | bash if bun is missing, which fetches and immediately executes remote code during runtime (https://bun.sh/install).