The Agent Skills Directory

[SAFE]: The skill implements a comprehensive and secure multi-agent coordination protocol. It features a strict 'Task contract' that centralizes control in the root agent and explicitly forbids sub-agents from spawning their own children or calling tasks, thereby preventing uncontrolled recursion and potential autonomy abuse. Security is further bolstered by the mandatory use of isolated workspaces for implementation tasks and the separation of roles, including a dedicated 'security-reviewer'. Crucially, the protocol requires that the root agent independently verify all sub-agent artifacts and reports, adhering to the principle that a worker's report cannot serve as its own verification. The instructions are transparent, consistent with the stated purpose, and include no obfuscation or suspicious external dependencies.

pstack-omp