Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution (prefercmd/memory as knowledge allow/deny lists), so repeated tasks start from proven commands and save tokens. Unified subagent routing and evolution: prefercmd/memory serve as knowledge allow/deny lists, saving tokens.
Skill Claude CodeCodex
Choose an existing evo-subagent role and prepare a focused delegation task. Use when deciding which agentkey to call or when a child task needs clearer scope and acceptance criteria.
Skill Claude CodeCodex
Review and curate an evo-subagent role's prefercmd.md and memory.md using evidence from actual runs. Use when evolution entries are stale, duplicated, contradictory, or too noisy.
Skill Claude CodeCodex
Diagnose evo-subagent provider/model routing and workspace binding failures in DeepSeek Harness. Use when a child runs on the wrong model, inherits unexpectedly, or fails before starting.