cpo-advisor
Pass
Audited by Gen Agent Trust Hub on Apr 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Python scripts (
scripts/pmf_scorer.pyandscripts/portfolio_analyzer.py) to calculate product-market fit scores and analyze product portfolios. These scripts are intended to process data but were not provided in the current file set for direct code analysis. - [PROMPT_INJECTION]: The 'Multi-Agent Analysis' section defines an orchestration framework that interpolates untrusted data (such as
{company},{product_list}, and{d30_retention}) into prompts for sub-agents. This creates a surface for indirect prompt injection if the source data (likecompany-context.mdor user input) contains malicious instructions. - Ingestion points: User-provided metrics (NPS, retention), product lists, and the
company-context.mdfile. - Boundary markers: None detected; variables are interpolated directly into the task prompt strings.
- Capability inventory: Execution of Python subprocesses (
pmf_scorer.py,portfolio_analyzer.py) and multi-agent task dispatching. - Sanitization: No explicit sanitization or validation of the interpolated variables is mentioned in the prompt templates.
Audit Metadata