vpe-advisor
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external JSON data files (such as
eng_state.json,team_metrics.json, andcapacity_inputs.json) to generate engineering organization health scores, productivity dashboards, and capacity plans. This represents a potential indirect prompt injection surface as the agent interprets the output of these scripts. However, this is an inherent part of the skill's primary functionality. - Ingestion points: Data is ingested from user-provided JSON files in the
vpe-advisor/scripts/execution workflows. - Boundary markers: The skill does not currently employ explicit boundary markers or instructions to the agent to ignore potentially malicious content within the input data files.
- Capability inventory: The skill executes local Python scripts and reads a local context file (
~/.claude/company-context.md). It does not perform network operations or access sensitive system credentials. - Sanitization: The Python scripts perform data validation and type casting but do not specifically sanitize string inputs for potential prompt injection patterns.
Audit Metadata