software-factory-mode-2026hashi
Warn
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: MEDIUMPROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions explicitly designed to override existing persistent guidelines found in files like CLAUDE.md and AGENTS.md. The description characterizes the skill as a mode with "brakes significantly removed," which indicates an intent to bypass established operational constraints or safety protocols.
- [PRIVILEGE_ESCALATION]: The instructions mandate that the agent perform sensitive and impactful actions without human oversight or confirmation. This includes pushing code changes directly to a repository in response to CI failures or review findings, posting inline review comments without checking the method, and automatically changing the status of pull requests from draft to "ready for review." These instructions increase the agent's operational autonomy and bypass standard "human-in-the-loop" safeguards for code management.
Audit Metadata