The Agent Skills Directory

[SAFE]: The skill defines a set of architectural principles for routing tasks to specialized agents without compromising security. Key defensive features include: 1) Isolated workspaces for parallel operations to prevent write conflicts. 2) Strict file ownership to maintain data integrity. 3) A verification model where worker reports are treated as evidence to be audited by the parent agent rather than automatically trusted. The instructions explicitly warn against using delegation to circumvent permission or sandbox boundaries.

model-orchestration