context-management

Context preservation across multi-agent workflows requires capturing decisions and state, then distributing tailored context to each agent. This Skill coordinates that capture, distribution, and checkpointing so each agent receives only relevant state without redundancy.

Long-running tasks need persistent checkpoints that capture key decisions and rationale at transition points. Checkpoints let you resume workflows without losing context and enable structured handoffs between agents or sessions.

Context compression filters and indexes past decisions to keep only what matters for downstream agents. Indexing enables fast retrieval of relevant decisions, while compression reduces token overhead in long conversations.

Yes. Agent-specific context distribution tailors the captured state for each agent's role and task. This reduces noise, improves focus, and ensures each agent sees only decisions and state relevant to its responsibilities.

Checkpoints are essential when workflows span multiple sessions, involve many agents, or accumulate long histories. They preserve coherent state efficiently, prevent context loss during handoffs, and maintain decision rationale without storing raw conversations.

Structured handoffs use checkpoints to surface relevant context at transition points. By capturing decisions and preparing agent-specific context before handoff, you maintain coherent state and prevent information gaps.