astoreyai/ww-context

WW Context Skill

Build comprehensive context from World Weaver memories tailored to the current task, project, and working directory.

Purpose

This skill synthesizes memories from all three subsystems (episodic, semantic, procedural) into actionable context for Claude. Unlike raw retrieval, it:

• Prioritizes relevance to current work
• Removes redundancy
• Highlights applicable skills
• Identifies knowledge gaps

When to Use

Invoke this skill when:

• Starting work on a task and need background
• User asks "what do we know about X?"
• Context seems missing from conversation
• Switching between projects
• Before making significant decisions

MCP Tools Used

mcp__ww-memory__recall_episodes    - Recent relevant episodes
mcp__ww-memory__semantic_recall    - Related entities
mcp__ww-memory__spread_activation  - Entity connections
mcp__ww-memory__recall_skill       - Applicable procedures
mcp__ww-memory__memory_stats       - System metrics

Context Building Workflow

Step 1: Gather Environmental Context

# Current directory
pwd

# Project identification
basename $(pwd)
git remote get-url origin 2>/dev/null || echo "Not a git repo"

# Recent activity
git log --oneline -5 2>/dev/null
git status --short 2>/dev/null

Step 2: Query Memory Systems

Episodic Query:

mcp__ww-memory__recall_episodes(
  query="[project name] [current task keywords]",
  limit=10,
  time_filter={after: "7 days ago"}
)

Semantic Query:

mcp__ww-memory__semantic_recall(
  query="[project name] [topic]",
  limit=15,
  include_relationships=true
)

Skill Query:

mcp__ww-memory__recall_skill(
  query="how to [task] in [project]",
  limit=5,
  check_preconditions=true,
  context={project, cwd}
)

Step 3: Synthesize Context

Combine results into structured context:

## Memory Context for [Task/Topic]

### Project: [Name]
- Directory: [path]
- Last activity: [date]
- Current state: [from git status]

### Relevant History
[Summarized episodes - what happened before]
- [Episode 1 summary]
- [Episode 2 summary]

### Key Knowledge
[Entities and their relationships]
- **[Entity A]**: [summary] → connected to [B, C]
- **[Entity B]**: [summary]

### Applicable Skills
[Procedures that match current context]
1. **[Skill name]**: [what it does]
   - Preconditions: [met/not met]
   - Steps: [brief outline]

### Gaps
[What's missing from memory]
- No recent episodes about [X]
- No skill for [Y]

### Recommendations
[Suggested actions based on context]

Step 4: Prioritize and Filter

Apply these filters:

1. Recency: Weight recent episodes higher
2. Relevance: Score against current task keywords
3. Diversity: Include different memory types
4. Actionability: Highlight immediately useful info

Output Formats

Brief Context (for quick reference)

**Context**: Working on [project], [task].
**Recent**: [1-2 relevant episodes]
**Know**: [2-3 key entities]
**Can use**: [1-2 skills]

Standard Context (default)

## Context: [Topic]

### Recent Work (3-5 episodes)
- [Episode summaries with dates]

### Key Knowledge (5-10 entities)
- [Entity name]: [summary]

### Skills (2-3 applicable)
- [Skill]: [brief description]

### Suggestions
- [Based on patterns in memory]

Deep Context (for complex tasks)

## Deep Context: [Topic]

### Timeline
[Chronological narrative of relevant work]

### Knowledge Graph
[Entity relationships visualized]

[A] --uses--> [B] [A] --relates--> [C]

### Full Skill Details
[Complete procedure for top skill]

### Pattern Analysis
[Recurring themes from episodes]

### Open Questions
[Uncertainties and gaps]

Context Quality Guidelines

Include

• Directly relevant episodes (score > 0.7)
• Entities connected to current project
• Skills matching preconditions
• Recent decisions and their outcomes

Exclude

• Old episodes (> 30 days) unless highly relevant
• Entities with no project connection
• Skills with unmet preconditions
• Redundant/duplicate information

Highlight

• Failures and lessons learned
• Successful patterns
• Unfinished work from previous sessions
• High-importance events

Examples

Example 1: Starting Work on Testing

Context request: "Build context for improving test coverage"

Output:
## Context: Test Coverage Improvement

### Recent Work
- [Nov 27] Fixed UUID import in benchmark test - success
- [Nov 26] Added pytest markers for HDBSCAN tests - success
- [Nov 25] Implemented session isolation tests - success

### Key Knowledge
- **pytest**: Test framework for Python
  - Connected to: coverage, markers, fixtures
- **Session Isolation**: Testing pattern for multi-tenant
  - Related: Neo4j, Qdrant, session_id

### Applicable Skills
1. **run-ww-tests**: Execute test suite
   - Steps: activate venv → pytest tests/ -v
   - Preconditions: Met (in WW directory)

2. **add-test-markers**: Mark tests for selective running
   - Steps: Add @pytest.mark decorator
   - Preconditions: Met

### Suggestions
- Current coverage: 79% - consider targeting uncovered modules
- Recent focus: integration tests - unit tests may need attention

Example 2: Debugging Issue

Context request: "Build context for Neo4j connection issues"

Output:
## Context: Neo4j Connection Debugging

### Recent Work
- [Nov 26] Fixed batch query parameter - changed $id to id
- [Nov 25] Investigated N+1 query pattern
- [Nov 24] Added connection pooling (50 connections)

### Key Knowledge
- **Neo4j**: Graph database, bolt://localhost:7687
  - Related: Cypher, Connection Pooling, Batch Queries
- **Connection Pooling**: 50 max connections
  - Config: pool_size in neo4j_store.py

### Applicable Skills
1. **debug-neo4j**: Check Neo4j connectivity
   - Test: curl http://localhost:7474
   - Logs: docker logs neo4j

### Suggestions
- Recent $id bug suggests parameter naming issues
- Check if variable vs parameter in Cypher
- Verify pool not exhausted

Integration

This skill is called by:

• /ww-context command
• ww-synthesizer agent
• SessionStart hook (for initial context)
• Other skills needing memory context