executing-phishing-simulation-campaign

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (critical risk: 1.00). The content is an explicit, operational guide for building phishing infrastructure and harvesting credentials (including tools/techniques to bypass MFA), i.e., clear instructions that enable credential theft and misuse.

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs collecting and processing captured credentials and even testing them against real systems (if authorized), which requires the LLM/agent to handle plaintext secrets and could lead to verbatim secret exposure.

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

• Third-party content exposure detected (medium risk: 0.30). The agent reads GoPhish campaign results from an external service endpoint at runtime (GET /api/campaigns/{id}/results via GoPhishClient.get_campaign_results), and if outsiders can author/submit email or landing-page contents that flow into those results, that free text can be ingested by the LLM-driven workflow for reporting.