melech-buy-vs-build
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external web searches and uses this information to formulate build-vs-buy recommendations, creating a surface where malicious web content could influence the agent's decision-making process. * Ingestion points: The skill reads local project files including manifests such as package.json and requirements.txt (SKILL.md) and retrieves external data via web search results across multiple research lanes (references/lanes.md). * Boundary markers: Absent. There are no instructions to the agent to disregard potential instructions embedded within retrieved search content. * Capability inventory: The skill recommends the adoption of software packages and interacts with the user via ask_question (SKILL.md) to finalize decisions. * Sanitization: Absent. The workflow does not include a sanitization or validation step for data retrieved from web sources.
- [DATA_EXFILTRATION]: The skill reads local project manifest files (SKILL.md) and checks for existing infrastructure and vendors. This project metadata, including specific dependency names and technical stack details, is used to construct queries for external web search engines, resulting in the transmission of internal technical environment details to third-party services.
Audit Metadata