StateMaster - State Management & Context Preservation
Overview
StateMaster manages state and context preservation for PromptForge AI. In multi-agent workflows, state includes conversation history, decisions made, partial results, and workflow progress. Losing state means losing work and forcing agents to redo effort.
When to Use
- When preserving context across agent handoffs or session boundaries
- When implementing state recovery after agent failures or restarts
- When synchronizing state between parallel agents working on the same task
- When managing long-running workflows that span multiple sessions
- When debugging state-related issues (lost context, stale data, inconsistent state)
- Don't use when: Working with stateless, single-turn interactions
Core Procedures
Step 1: Identify State Components
For each workflow, identify what state must be preserved:
- Conversation State: Message history, turn count, active topics
- Decision State: Choices made, alternatives considered, rationale
- Artifact State: Files, prompts, outputs generated so far
- Workflow State: Current step, completed steps, pending steps
- Agent State: Each agent's current task, progress, blockers
Step 2: Design State Storage
Choose appropriate storage for each state type:
- Short-term (in-session): In-memory state, fast access, lost on restart
- Medium-term (cross-session): Database or file storage, survives restarts
- Long-term (persistent): Knowledge base, indexed and searchable
- Shared State: Accessible by all agents in the workflow
- Private State: Only accessible by the owning agent
Step 3: Implement State Synchronization
When multiple agents share state:
- Locking: Prevent concurrent writes to the same state object
- Versioning: Track state versions to detect conflicts
- Merge Strategy: Define how to resolve conflicting state updates
- Notification: Alert agents when shared state changes
- Consistency Check: Periodically verify all agents have consistent state
Step 4: Implement State Recovery
For fault tolerance:
- Checkpointing: Save state at key workflow milestones
- Recovery Point: On failure, restore to last successful checkpoint
- State Validation: After recovery, verify state is complete and consistent
- Replay Capability: If needed, replay actions from checkpoint to current state
Step 5: Monitor State Health
- State size growth rate (detect unbounded growth)
- State access latency (slow state access slows workflows)
- State conflict rate (how often concurrent writes collide)
- Recovery success rate (how often recovery restores valid state)
- Stale state detection (state that hasn't been updated but should be)
Success Metrics
- State loss rate: 0% (no state lost during normal operations)
- State recovery time: <30 seconds from failure to restored state
- State conflict rate: <1% of state updates
- State access latency: <100ms for 95% of accesses
- Checkpoint frequency: At least every 5 workflow steps
Error Handling
- Error: State corruption detected (invalid or inconsistent state)
Response: Restore from last valid checkpoint, log corruption details, investigate root cause
- Error: State size exceeds storage limits
Response: Archive old state data, compress current state, alert if growth is abnormal
- Error: Agent reads stale state (state changed but agent has old version)
Response: Invalidate agent's state cache, force refresh from current state, add version checking
Cross-Team Integration
Related Skills: flowdesigner-promptforge-agent-handoff, cascade-promptforge-workflow-designer, archivist-promptforge-knowledge-management, memory-system-management
Used By: Cascade, FlowDesigner, Harmonic, ALL agents that maintain state across interactions
1---2name: statemaster-promptforge-state-management3description: Use when managing conversation state, context preservation across sessions, memory synchronization, and state recovery for multi-agent workflows. This skill ensures no context is lost between agent interactions or across session boundaries.4---56# StateMaster - State Management & Context Preservation78## Overview9StateMaster manages state and context preservation for PromptForge AI. In multi-agent workflows, state includes conversation history, decisions made, partial results, and workflow progress. Losing state means losing work and forcing agents to redo effort.1011## When to Use12- When preserving context across agent handoffs or session boundaries13- When implementing state recovery after agent failures or restarts14- When synchronizing state between parallel agents working on the same task15- When managing long-running workflows that span multiple sessions16- When debugging state-related issues (lost context, stale data, inconsistent state)17- **Don't use when:** Working with stateless, single-turn interactions1819## Core Procedures2021### Step 1: Identify State Components22For each workflow, identify what state must be preserved:23- **Conversation State:** Message history, turn count, active topics24- **Decision State:** Choices made, alternatives considered, rationale25- **Artifact State:** Files, prompts, outputs generated so far26- **Workflow State:** Current step, completed steps, pending steps27- **Agent State:** Each agent's current task, progress, blockers2829### Step 2: Design State Storage30Choose appropriate storage for each state type:31- **Short-term (in-session):** In-memory state, fast access, lost on restart32- **Medium-term (cross-session):** Database or file storage, survives restarts33- **Long-term (persistent):** Knowledge base, indexed and searchable34- **Shared State:** Accessible by all agents in the workflow35- **Private State:** Only accessible by the owning agent3637### Step 3: Implement State Synchronization38When multiple agents share state:39- **Locking:** Prevent concurrent writes to the same state object40- **Versioning:** Track state versions to detect conflicts41- **Merge Strategy:** Define how to resolve conflicting state updates42- **Notification:** Alert agents when shared state changes43- **Consistency Check:** Periodically verify all agents have consistent state4445### Step 4: Implement State Recovery46For fault tolerance:47- **Checkpointing:** Save state at key workflow milestones48- **Recovery Point:** On failure, restore to last successful checkpoint49- **State Validation:** After recovery, verify state is complete and consistent50- **Replay Capability:** If needed, replay actions from checkpoint to current state5152### Step 5: Monitor State Health53- State size growth rate (detect unbounded growth)54- State access latency (slow state access slows workflows)55- State conflict rate (how often concurrent writes collide)56- Recovery success rate (how often recovery restores valid state)57- Stale state detection (state that hasn't been updated but should be)5859## Success Metrics60- State loss rate: 0% (no state lost during normal operations)61- State recovery time: <30 seconds from failure to restored state62- State conflict rate: <1% of state updates63- State access latency: <100ms for 95% of accesses64- Checkpoint frequency: At least every 5 workflow steps6566## Error Handling67- **Error:** State corruption detected (invalid or inconsistent state)68 **Response:** Restore from last valid checkpoint, log corruption details, investigate root cause69- **Error:** State size exceeds storage limits70 **Response:** Archive old state data, compress current state, alert if growth is abnormal71- **Error:** Agent reads stale state (state changed but agent has old version)72 **Response:** Invalidate agent's state cache, force refresh from current state, add version checking7374## Cross-Team Integration75**Related Skills:** flowdesigner-promptforge-agent-handoff, cascade-promptforge-workflow-designer, archivist-promptforge-knowledge-management, memory-system-management76**Used By:** Cascade, FlowDesigner, Harmonic, ALL agents that maintain state across interactions
Construct-AI-primary