overseer_skill

Overview

This skill manages multi-session work and task handoffs using the Overseer codemode MCP. It treats work as structured tickets (description, context, result) and persists full context across sessions so teams can pick up work, record decisions, and verify outcomes. Use it to break down complex work into a three-level hierarchy: milestone → task → subtask.

How this skill works

Overseer stores tasks in a SQLite-backed database and exposes methods to fetch the next ready task with full inherited context or list ready tasks for status checks. Workflows use nextReady() to obtain a deepest ready leaf (TaskWithContext), start() to create a work bookmark, and complete() to record results and bubble learnings up the chain. Blockers, hierarchy constraints, and explicit verification are enforced to keep progress reliable.

When to use it

• When work spans multiple sessions and context must persist between handoffs
• When breaking complex features into tasks and subtasks with clear ownership
• When you need a verifiable trail of decisions and implementation results
• When recording learnings and bubbling them to parent milestones is valuable
• When persistence and structured hierarchy are required instead of ephemeral todos

Best practices

• Right-size tasks so each is completable in a single focused session
• Write action-oriented descriptions that start with a verb and state clear done criteria
• Use nextReady() at the start of a work session to get full context and inherited learnings
• Record explicit verification evidence in the result when completing a task
• Avoid referencing ephemeral task IDs in external artifacts; describe the work itself
• Limit children to about 3–7 per parent to keep decomposition practical

Example use cases

• Implementing a feature that requires design, backend, and frontend steps across days
• Handing off a partially completed bugfix to another engineer with full context
• Running a focused work session: fetch nextReady(), start(), implement, complete() with verification
• Recording architectural decisions and follow-up tech-debt tasks after completing a change
• Coordinating multi-step rollout tied to a milestone with several dependent tasks