sharecrm

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes several Python scripts (e.g., prepare_create_form.py, query_bom_batch.py, query_object_data_file.py) that wrap the local sharecrm CLI. These scripts use subprocess.run to execute commands with structured JSON arguments. The approach is controlled and used for data processing tasks like form preparation and price calculations.
  • [EXTERNAL_DOWNLOADS]: The skill references an external JSON file (https://a9.fspage.com/FSR/weex/sfa/resource/map_boundary.json) for map visualization data. This resource is hosted on a vendor-owned domain (fspage.com) and is used for rendering regional distribution charts in HTML reports.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from ShareCRM, including meeting summaries, customer feedback, and record details. While this represents a surface for indirect prompt injection from external records, the skill processes this information into structured reports or uses it as parameters for further CLI calls rather than direct instruction execution.
  • [DYNAMIC_EXECUTION]: Helper scripts such as query_attribute_relation_batch.py and query_bom_batch.py dynamically construct SQL queries for the ShareCRM CLI. These scripts include sanitization logic (e.g., sql_quote functions) to escape single quotes in user-supplied values, mitigating basic injection risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:14 PM