The Agent Skills Directory

[SAFE]: The skill mandates an isolated execution environment for sub-agents, where they receive only the specific context needed for their task rather than the full session history or implementation plans. This isolation prevents context contamination and reduces the exposure of broader project metadata to individual sub-tasks.
[SAFE]: A mandatory double-stage review process involving specification consistency and code quality is enforced for every task. This structure ensures that all sub-agent output is verified against original requirements and security standards before being integrated.
[SAFE]: The skill includes explicit safety constraints, such as forbidding implementation on main or master branches without explicit user approval, which protects against unintended or destructive code changes.

subagent-driven-development