skills/mukul975/anthropic-cybersecurity-skills/implementing-endpoint-dlp-controls/Gen Agent Trust Hub
implementing-endpoint-dlp-controls
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local directories, event logs (USB and Cloud), and CSV exports for analysis.\n
- Ingestion points:
scripts/agent.py(via--scan-dir,--usb-log,--cloud-log) andscripts/process.py(via CSV input).\n - Boundary markers: None found; the scripts process file content and logs directly without specific delimiters or instructions to ignore embedded commands.\n
- Capability inventory: The scripts perform file reads, regex pattern matching, and JSON/CSV parsing. No network exfiltration or shell command execution is triggered by the parsed data.\n
- Sanitization: The scripts use standard libraries with basic error handling but do not specifically sanitize content against prompt injection if the resulting reports are later processed by an LLM.\n- [EXTERNAL_DOWNLOADS]: The documentation provides implementation examples for external API interaction with well-known security services.\n
- Evidence:
references/api-reference.mdcontains code snippets for querying Microsoft Graph and CrowdStrike APIs.
Audit Metadata