Prompts

Hive includes 4 MCP prompts — structured protocols that any MCP client can invoke to follow multi-step workflows.

retrospective

End-of-session review that extracts lessons and appends them to the vault.

Parameters: project (string)

Protocol:

1. Review work completed in the current session
2. Identify patterns, mistakes, and insights
3. Format as structured lessons
4. Use vault_write to append to the project’s 90-lessons.md

Usage: Ask your assistant to “run a retrospective for my-project” at the end of a work session.

delegate

Structured protocol for delegating tasks to cheaper models via hive-worker.

Parameters: task (string)

Protocol (as the prompt states it today):

1. Suitability check against a delegatable / not-delegatable matrix
2. Budget check: worker_status for remaining budget and model availability
3. Context compression: summarize relevant vault files and strip the task to its essential instruction
4. Delegate: pick a tier with worker_status, then call delegate_task, one task per call
5. Evaluate the result, and handle the task directly if it is poor

The worker has no budget or model tiers any more; the prompt’s wording catches up in #520, and this page changes with it.

Usage: Ask your assistant to “delegate this task: explain this regex”

vault_sync

Post-sprint vault synchronization — reconcile documentation with shipped code.

Parameters: project (string)

Protocol:

1. Load current project context and tasks
2. Compare with recent git history
3. Identify stale docs, completed tasks, missing documentation
4. Update vault files to match current state

Usage: Ask your assistant to “sync the vault for my-project after this sprint”

benchmark

Estimate token savings from hive MCP tools in the current session.

Parameters: none

Protocol:

1. Call vault_health(include_usage=True) to get tool call statistics
2. Estimate tokens that would have been consumed by static loading
3. Calculate savings percentage
4. Report results

Usage: Ask your assistant to “benchmark the token savings from hive”