investigating-agentforce-d360
Pass
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill interacts with the Salesforce
sfCLI usingsubprocess.runto manage authentication and retrieve organization metadata (scripts/dc.py,scripts/fetch_dc.py). - Arguments are passed as lists to prevent shell injection.
- User-supplied inputs such as organization aliases are used within defined CLI flags.
- [DATA_EXFILTRATION]: The skill retrieves session logs and conversation transcripts from Salesforce Data Cloud and writes them to local JSON and Markdown files (
scripts/storage.py,scripts/fetch_dc.py). - This behavior is consistent with the skill's stated purpose of session investigation.
- Network requests are directed to the authenticated Salesforce instance URL obtained from the official CLI.
- [SAFE]: The skill implements a centralized validation framework (
scripts/_shared/fs_guard.pyandscripts/_shared/paths.py) that uses anchored regular expressions to sanitize all session IDs, API names, and organization IDs. - This prevents path traversal vulnerabilities by ensuring all components of a filesystem path are strictly validated before use.
- The skill is authored by 'forcedotcom' (Salesforce), and its resource patterns align with the vendor's legitimate infrastructure.
Audit Metadata