Manage tasks, state, and memories - parse PRDs into hierarchical tasks with dependencies and complexity
The taskmanager skill is a structured task and state management system built around a SQLite database that serves as the single source of truth for all project tasks, milestones, analyses, and memories. It is designed to transform product requirements (PRDs) into a clear, hierarchical task tree with explicit dependencies and manageable complexity. Instead of relying on informal planning or scattered notes, it enforces a disciplined, database-driven workflow where every task, state change, and memory is tracked and persisted. This ensures consistency, traceability, and integrity across planning and execution phases.
The skill provides strict level-by-level task expansion. When given PRD content—either from a markdown file or a direct prompt—it first generates top-level tasks (conceptual epics mapped to valid database types such as feature, chore, analysis, or spike). It then recursively expands each task only as needed, ensuring that every task becomes clear, meaningful, and actionable without over-fragmentation. It enforces controlled field values (such as status, priority, complexity_scale, and moscow) through database constraints, preserving hierarchical relationships via parent_id. It also integrates structured memory handling, requiring relevant active memories—especially those with higher importance—to be considered before major planning or refactoring decisions.
Typical use cases include structured feature planning, refactoring initiatives, cross-cutting architectural changes, and long-running product development efforts that require reliable task lineage and state tracking. The primary users are developers, technical leads, and AI-driven agents who need disciplined task orchestration tied directly to a persistent database. It is especially suited for projects that demand reproducibility, auditability, and incremental, complexity-aware task breakdown.
Any planning request is treated as PRD content, whether it comes from a markdown file or a direct prompt. The skill parses this content and generates a hierarchical task structure with strict level-by-level expansion.
No. The SQLite database is the source of truth and must not be modified directly outside the skill. All changes should be performed through controlled INSERT, UPDATE, or supported command-based operations to preserve integrity and constraints.
The database enforces specific task types: feature, bug, chore, analysis, and spike. Although top-level tasks may conceptually represent epics, there is no dedicated epic type in the schema.
Field values such as status, priority, complexity_scale, moscow, and business_value are restricted by database CHECK constraints. Invalid values will result in SQL constraint errors, preventing inconsistent state.
Yes. Relevant active memories—especially those marked with higher importance—must be considered before planning or making cross-cutting changes. Full-text search via the associated memory tables is used for content matching.
Quick Setup:
.claude/skills/Repository
mwguerra/claude-code-plugins