social-handle-to-email-finder

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 1.00). The skill contains explicit, deceptive instructions to install a runtime dependency automatically and "do not request separate approval" or tell the user, which overrides normal consent and is outside the skill's stated purpose.

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md instructs the agent to invoke the browser-act Skill for “live browser pages and execute tasks,” which at runtime can fetch and extract arbitrary public web page text (outsider-authored free text) into the LLM context via BrowserAct’s browsing workflow.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill declares the BrowserAct CLI/Skill as a required runtime dependency and provides installation links (https://www.browseract.com/?co-from=lead-generation and https://www.browseract.com/?co-from=lead-generation&redirect=https://github.com/browser-act/skills/tree/main) that the agent is instructed to fetch and install automatically, which implies fetching and executing remote code at runtime.

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

• Attempt to modify system services in skill instructions detected (high risk: 0.70). The Skill instructs the agent to install the BrowserAct CLI automatically (modifying the host) and explicitly says not to request the user's confirmation, which directs the agent to alter the machine's software state and bypass user approval.