soia-dev-test-draft-doc
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides installation steps that fetch software from remote sources, specifically
github.com/soia-team/soia-open-skillsandgithub.com/soia-team/soia-open-dev-skills. These are managed via plugin marketplaces and thenpxutility. - [COMMAND_EXECUTION]: Instructions are provided for users to execute shell commands, including
npx skills addandclaude plugin install. These commands include flags for global installation (-g) and automatic confirmation (-y), which modify the user's environment or agent configuration. - [PROMPT_INJECTION]: The skill is designed to ingest external requirements documents from user-provided text or public URLs. This presents a surface for indirect prompt injection, where an attacker could embed malicious instructions within a PRD to influence the agent's behavior. The skill contains defensive instructions to define input boundaries and verify requirements to reduce this risk.
- [DATA_EXFILTRATION]: The skill processes potentially sensitive business requirements and PRDs. Although it includes rules for data sanitization, placeholders, and avoiding unauthorized file access, the processing of external materials remains a security consideration for data exposure.
Audit Metadata