Context Building

Context Building

Purpose / When to Activate

Activate when context is fragmented or multiple memory sources need to be merged before a complex task:

• After memory-retrieve when inputs come from multiple memory files
• Before planning or implementation skills on complex, multi-artefact Stories
• When previous context bundles are stale or need restructuring

For simple, single-artefact tasks, an agent may reason directly without this skill. This skill does not retrieve knowledge — it structures and composes already-selected inputs into a focused bundle.

Process

1. Validate inputs are scoped and minimal

2. Remove duplicated or overlapping information

3. Prioritize: acceptance criteria, constraints, decisions, current artefacts

4. Structure the bundle in canonical order:

   • Current Objective
   • Governed Artefacts (Epics / Stories / Plans)
   • Acceptance Criteria & Constraints
   • Applicable Rules
   • Relevant Memory

5. Enforce minimal token footprint

Outputs

=== EXECUTION CONTEXT ===

OBJECTIVE:
...

ARTEFACTS:
...

ACCEPTANCE & CONSTRAINTS:
...

RULES:
...

RELEVANT MEMORY:
...

=========================

No prose. No fluff. No hidden reasoning.

Quality Checks

• Context is focused and task-specific
• No irrelevant memory included
• No artefacts outside scope
• Token usage minimized
• All constraints explicit

Non-Goals

This skill must NOT:

• Retrieve memory (use memory-retrieve first)
• Infer missing artefacts
• Add assumptions or expand scope
• Inject creative interpretation

Retrieval decides what is relevant. Context building decides how it is fed to reasoning.