Transform CHANGE_REQUEST and PRD into a dependency-aware DAG execution plan.
npx skills add https://github.com/kinetas/harness_engineering --skill changestart
Please help me install this Agent Skill. Skill: changeStart Source: https://github.com/kinetas/harness_engineering/tree/main/changeStart Command: npx skills add https://github.com/kinetas/harness_engineering --skill changestart
It turns an incoming change request into an actionable, dependency-aware development plan that can run in parallel without duplicating work or stepping outside the approved scope.
doc/company_state.json task counters and checks collectorMode and monitoringEnabled to decide monitoring and reporting behavior.Use case: You receive a CHANGE_REQUEST describing multiple fixes and improvements, and you want the system to partition the work into parallel segments, respect file ownership boundaries, and keep report/report.md continuously updated.
Run /changeStart to generate a DAG, split the change into dependency islands, delegate each segment to a Manager AI, and update report/report.md as the segments complete.
High-intent search queries and answers about installing and using this skill.
To parallelize multi-file code changes safely, you can build a dependency-aware DAG that clusters tasks into isolated segments based on file ownership. This ensures parallel execution respects boundaries and prevents overlapping modifications.
Dependency-island segmentation is a change management technique that groups tasks touching the same files into isolated clusters. It separates verification testing and shared infrastructure to enable safe parallel execution without duplicated work.
You can generate a topological execution plan by reading a CHANGE_REQUEST document and building a dependency graph. The system layers tasks to determine execution order and partitions them into parallel segments managed by spawned AI agents.
Yes, you can limit spawned AI agents during parallel task orchestration by configuring manager and sub-agent limits. The system checks these boundaries before delegating dependency-island segments to Manager AIs.
DAG orchestration tracks progress for parallel development tasks by updating a centralized state file and appending completion statuses to a consolidated report. A Collector AI can optionally monitor and update this report as segments finish.
Yes, a PRD is required alongside the change request to build a dependency-aware execution plan. The system reads both documents to construct the DAG, apply topological layering, and segment tasks into isolated file-ownership clusters.
kinetasConfigure Manager and Sub AI concurrency limits in workspace files.
kinetasTransforms PRD tasks into a dependency DAG with partitioned segments for parallel execution.
curiositechPlan and optimize DAG execution graphs for multi-step tasks.
tmalcolm-0607Orchestrate wave-based parallel feature implementation across isolated git worktrees.