Context Engineering Skill
This skill provides techniques for managing context effectively throughout the workflow to prevent bloat and maintain efficiency.
Core Principles
-
Stay Under 40%: Orchestrator should never exceed 40% context usage
-
Fresh Context Per Task: Each major task gets a fresh agent with clean context
-
Load Only What's Needed: Don't load full files when summaries suffice
-
Delegate Deep Work: Spawn agents for context-heavy operations
Context Budget Guidelines
Role Context Budget Reason
Orchestrator <40% Needs room for coordination
Interviewer <70% Conversation history accumulates
Planner <60% Needs to see requirements + codebase patterns
Executor <80% Can focus on single task deeply
Verifier <60% Needs to cross-reference requirements and code
Context Loading Strategies
Selective Loading
Instead of loading full files:
BAD: Read entire codebase GOOD: Read specific files needed for current task BEST: Read summaries, then specific sections
Hierarchical Loading
Load context in layers:
-
L1 Summary: High-level overview (FLOW.md)
-
L2 Details: Item-specific state (ITEM-XXX.md)
-
L3 Specifics: Task-specific files (only files in current task)
Incremental Loading
Load more context only when needed:
- Start with minimal context
- Identify what's missing
- Load specific missing pieces
- Repeat as needed
Fresh Context Patterns
When to Use Fresh Context
Spawn a fresh agent when:
-
Starting a new task in EXECUTE phase
-
Moving to a new phase
-
Current context is getting heavy
-
Deep codebase exploration needed
Fresh Context Handoff
When spawning fresh agent:
Include in handoff:
- Task ID and definition
- Key decisions from ITEM-XXX.md
- Relevant file paths
- Success criteria
Exclude:
- Full conversation history
- Files not related to task
- Exploration paths not taken
Context Reset Points
Natural points to reset context:
-
Phase transitions
-
After each completed task
-
After resolving a blocker
-
After user requests fresh start
Context Monitoring
Signs of Context Bloat
Watch for:
-
Repeated information in responses
-
Slower response times
-
Forgetting earlier context
-
Inconsistent references
Context Health Check
Periodically verify:
-
Can still reference key decisions?
-
Can identify current task?
-
Can access state files?
-
Response quality maintained?
State File as External Memory
Use state files as external memory:
What to Store in Files
-
All decisions (ITEM-XXX.md Decisions section)
-
Exploration progress (ITEM-XXX.md Exploration section)
-
Task definitions (ITEM-XXX.md Tasks section)
-
Requirements (ITEM-XXX.md Requirements section)
What to Keep in Context
-
Current phase and task
-
Key constraints
-
Immediate next steps
-
Recent decisions (can reference file for older ones)
Delegation Patterns
Deep Exploration Delegation
Task: Understand how authentication works in this codebase
Instead of:
- Loading all auth-related files into orchestrator context
Do:
- Spawn researcher agent with specific question
- Receive summary findings
- Keep summary in orchestrator context
Implementation Delegation
Task: Implement TASK-005
Instead of:
- Loading full codebase patterns and task into orchestrator
Do:
- Spawn executor with task definition + relevant files only
- Receive completion status
- Update state files
Context Compression Techniques
Summary Generation
After deep work, compress to summary:
Before compression: [Full exploration transcript]
After compression: Exploration Result: Authentication
- Uses JWT tokens
- Stored in localStorage
- Refresh flow in AuthService.refresh()
- Decision needed: token expiry duration
Reference Instead of Inline
BAD: Include full task XML in every message GOOD: "Executing TASK-005 (see ITEM-XXX.md Tasks section)"
Checkpoint Snapshots
At checkpoints, snapshot key state:
CHECKPOINT: EXECUTE-001
Context snapshot:
- Phase: EXECUTE
- Progress: 5/10 tasks
- Current: TASK-006
- Blocked: None
- Key decisions: Use JWT (DECISION-003)
Integration with Workflow
During DISCUSS
-
Keep exploration map in ITEM-XXX.md
-
Summarize findings per area
-
Reference decisions by ID
During PLAN
-
Read requirements once, create tasks
-
Don't reload requirements repeatedly
-
Reference task IDs not full definitions
During EXECUTE
-
Fresh context per task
-
Load only task + relevant files
-
Update state after each task
During VERIFY
-
Load summary of completed work
-
Cross-reference requirements by ID
-
Keep verification results in ITEM-XXX.md
Best Practices
-
Reference, don't repeat: Use IDs and file paths
-
Summarize immediately: After deep work, compress findings
-
Clean handoffs: Fresh agents get focused context
-
Use state files: They're your external memory
-
Monitor bloat: Watch for degradation signs
-
Delegate proactively: Don't wait until context is full