pricing-strategy
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (product details, cost data, and competitor info) to generate pricing strategies and simulations. Malicious instructions embedded in this input could potentially influence the behavior of the specialized agents during analysis.\n
- Ingestion points: Phase 1 extracts user input and saves it to
_workspace/00_input.md, which is then read by multiple agents.\n - Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore' instructions when storing or reading user data in the workspace.\n
- Capability inventory: The system involves file system access (creating directories, writing/reading markdown files), inter-agent messaging via
SendMessage, and web search capabilities for competitive research.\n - Sanitization: No explicit sanitization, validation, or escaping of external content is mentioned before the data is passed into the agent pipeline.\n- [COMMAND_EXECUTION]: The skill requires the orchestrator to perform file system operations, specifically creating a
_workspace/directory and organizing input/output files. These operations are restricted to the local project environment and support the primary function of the skill.
Audit Metadata