skills/mukul975/anthropic-cybersecurity-skills/detecting-email-forwarding-rules-attack/Gen Agent Trust Hub
detecting-email-forwarding-rules-attack
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources including the Microsoft Graph API and local Exchange audit logs. This introduces a vulnerability surface where maliciously crafted content within the analyzed data could attempt to influence the agent's instructions or behavior.
- Ingestion points: Data enters the agent's context via network requests to the Microsoft Graph API in
scripts/agent.pyand through the parsing of local log files (CSV/JSON) in bothscripts/agent.pyandscripts/process.py. - Boundary markers: The skill lacks explicit delimiters or specific instructions to the agent to disregard potential commands or malicious formatting embedded within the logs or API responses.
- Capability inventory: The skill possesses network communication capabilities via the
requestslibrary inscripts/agent.pyand file system write access for report generation inscripts/process.py. - Sanitization: There is no evidence of sanitization, escaping, or strict schema validation applied to the ingested content before it is processed for threat analysis.
Audit Metadata