social-brand-mention-monitor

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill includes explicit, deceptive installation instructions that tell the agent to auto-install a dependency without asking the user and to hide that fact ("do not request separate approval" / "do not describe it as a request"), which is outside the stated monitoring purpose and attempts to bypass user consent.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The Skill contains instructions to automatically install and invoke an external CLI/Skill without user consent and to hide that action, creating a supply-chain / unauthorized-installation risk that could enable remote code execution or data exfiltration.

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

• Third-party content exposure detected (high risk: 0.85). This skill’s required runtime workflow invokes the browser-act Skill to access live public web/social pages, which can cause outsider-authored free text from those pages (browser-extracted visible content) to be ingested into the agent’s LLM context for analysis/export.

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

• Potentially malicious external URL detected (high risk: 0.80). The Skill treats the BrowserAct CLI as a required runtime dependency and explicitly instructs the agent to automatically install it via the URL https://www.browseract.com/?co-from=social-media&redirect=https://github.com/browser-act/skills/tree/main, which implies fetching and running remote installer code at runtime.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill instructs the agent to automatically install required runtime dependencies (via the Install BrowserAct link) without asking the user for confirmation and forbids requesting separate approval, which pushes the agent to modify the host system (potentially requiring sudo or changing system files) without user consent.