What problem does it solve?
Turning codebase research, bug investigations, and competitor analysis into an actionable, executable backlog normally requires manual synthesis and task authoring. This Skill automates that pipeline: it dispatches parallel explorer agents, synthesizes findings, and emits a validated task graph that cy-execute-task can run directly.
Core Features & Use Cases
- Parallel Explorer Dispatch: Slices a research question into 4-8 scoped analyses run via
compozy exec with GLM 5.2 through the pi harness, each producing a seven-section evidence-cited analysis document.
- Lite Spec Authoring: Generates a
_spec.md with root causes, target architecture, delete targets, sequencing waves, and impact audits, without interactive ADR ceremony.
- Executable Task Graph: Produces
task_NN.md issue files and a _tasks.md compozy.tasks/v2 manifest with dependencies, complexity ratings, and Unit/Integration/E2E test plans, validated via compozy tasks validate.
- Use Case: When auditing a subsystem or investigating a bug class (e.g., "returning to a live session shows a blank thread"), run the skill with a slug and research prompt to receive a complete, validated backlog under
.compozy/tasks/<slug>/ ready for execution.
Quick Start
Ask the agent to research why returning to a live session shows a blank thread and turn the findings into an executable task suite under the slug session-reconnect.
