# Disruption Aware Reactive Replanning > Skill: disruption-aware-reactive-replanning - Skill: `dingxingdi/disruption-aware-reactive-replanning` (Agent Skill, multi-file: 5 files) - Install (CLI): `npx skillmds@latest add dingxingdi/disruption-aware-reactive-replanning` - Raw SKILL.md: https://api.skillmd.com/api/skills/dingxingdi/disruption-aware-reactive-replanning/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Coding & Dev Tools - Author: dingxingdi (https://skillmd.com/u/dingxingdi) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/dingxingdi/disruption-aware-reactive-replanning --- # Skill: disruption-aware-reactive-replanning ## 1. Capability Definition & Real Case * **Professional Definition**: Disruption-aware reactive replanning (encompassing event-driven context synchronization) is the orchestration capability to dynamically modify an active execution trace in response to exogenous environmental shifts, pushed telemetry streams, or resource unavailability that occurs after a plan has been initiated. Instead of relying solely on periodic static polling, the orchestrator incorporates live asynchronous events (e.g., a machine failure, a topological blockage, or a critical payload update) to immediately invalidate outdated assumptions in its dependency graph (DAG). The agent then performs a surgical state-update or re-routes the workflow to restore progress while preserving the unaffected portions of its global strategy. * **Dimension Hierarchy**: Trustworthy Execution->Verification and Resilience->disruption-aware-reactive-replanning ### Real Case **[Case 1]** * **Initial Environment**: A multi-vehicle urban transport environment is active with three vehicles and five passengers. The system has already matched passengers to vehicles and established pick-up routes. A road-closure map service is available as a background monitor pushing live events. * **Real Question**: A major bridge on the route to the airport has just been closed for emergency repairs. Update the ride-sharing schedule for all affected vehicles to ensure no passenger misses their flight, prioritizing the earliest departures. * **Real Trajectory**: The orchestrator receives the disruption alert mid-task and queries current GPS coordinates. Identifying that Vehicle 1 is 5 minutes from the bridge, it halts the 'Navigate' action, updates the DAG with a new 20-minute detour, and offloads Passenger 3 to Vehicle 2 (which is unaffected) to prevent cascading delays. * **Real Answer**: Ride-share schedule updated: Vehicle 1 rerouted via detour; Passenger 3 transferred to Vehicle 2 to preserve deadline. * **Why this demonstrates the capability**: Demonstrates reactive replanning because the disruption happened mid-execution, invalidating a previously optimized route. The agent performed state-aware filtering to selectively re-balance load rather than restarting from scratch. --- **[Case 2]** * **Initial Environment**: A production environment (Job-Shop) contains 3 machines and 3 active jobs. Job 1 is currently 50% complete on Machine A. Machines provide real-time status telemetry push-alerts. * **Real Question**: Machine A has experienced a mechanical failure and will be offline for the next 2 hours. Reschedule the remaining operations to minimize the final completion time (makespan), ensuring no job steps are skipped. * **Real Trajectory**: The agent observes the pushed 'Machine Breakdown' event at time T=4. It marks Machine A as blocked, identifies that Job 1 Step 2 can be moved to Machine B, and performs a 'Swap-and-Insert' operation in the workflow DAG to maintain overall operational precedence. * **Real Answer**: Schedule revised: Job 1 Step 2 shifted to Machine B, resuming operations immediately to minimize makespan impact. * **Why this demonstrates the capability**: Highlights the resilience aspect of workflow orchestration. The unexpected event forces a cascading logic change requiring alternatives to be evaluated dynamically to avert total systemic stall. --- **[Case 3]** * **Initial Environment**: A large-scale cloud resource management environment where 'Resource Watchers' push immediate asynchronous changes (deltas) into the agent's vector index rather than relying on hourly polling. * **Real Question**: What is the current headroom for the guaranteed-throughput slice? If it's below 10Mbps, increase it immediately. * **Real Trajectory**: While processing, the system detects a resource fluctuation via the Push Path, updating 'headroom' to 8Mbps. The orchestrator receives this up-to-the-second snapshot, recognizing the shortage mid-task. It executes a 'Resource Update' tool to modify the blueprint to 15Mbps, parsing the success based on the new watcher state. * **Real Answer**: The live headroom alert showed 8Mbps due to a recent spike. I have increased throughput to 15Mbps to restore the safety threshold. * **Why this demonstrates the capability**: This illustrates the event-driven synchronization dimension of replanning. The orchestrator reacted strictly to a live push-update mid-session, surgically updating its execution trajectory. Failing to react to the disruption would have resulted in using stale polling data. ## Pipeline Execution Instructions To synthesize data for this capability, you must strictly follow a 3-phase pipeline. **Do not hallucinate steps.** Read the corresponding reference file for each phase sequentially: 1. **Phase 1: Environment Exploration** Read the exploration guidelines to discover raw knowledge seeds: `references/EXPLORATION.md` 2. **Phase 2: Trajectory Selection** Once Phase 1 is complete, read the selection criteria to evaluate the trajectory: `references/SELECTION.md` 3. **Phase 3: Data Synthesis** Once a trajectory passes Phase 2, read the synthesis instructions to generate the final data: `references/SYNTHESIS.md`