| name | creating-context-packs-for-ai |
| description | Creates structured context documents that AI systems can consume to generate consistent, accurate content for a feature. Use when preparing a feature for AI-assisted content generation, building AI writing workflows, or enabling LLMs to write within product constraints. |
Creating Context Packs for AI
Quick start
Collect or infer:
- Feature name and scope
- Existing context documents (state maps, vocabulary, content briefs)
- AI system requirements (token limits, format constraints)
- Content types the AI will generate
- Guardrails and constraints the AI must respect
Then produce output using TEMPLATES.md. Validate with RUBRIC.md.
Workflow
- Gather all existing feature context (vocabulary, states, constraints, tone).
- Identify what content the AI will generate and in what contexts.
- Structure context following the context pack schema.
- Include explicit constraints and prohibited patterns.
- Add examples of correct and incorrect outputs.
- Validate completeness using the completeness checklist.
- Test with target AI system to verify usability.
- Run the rubric check. Revise until it passes.
Degrees of freedom
- Default: Low. Context packs must be precise and complete.
- Allowed variation: Format may adapt to AI system requirements. Detail level may vary by content complexity.
Failure modes to avoid
- Omitting constraints that lead to off-brand content
- Including redundant information that wastes tokens
- Using ambiguous language that AI interprets differently than intended
- Missing state-specific guidance (AI doesn't know what state it's writing for)
- Providing examples without explaining why they're correct
- Creating context packs that exceed AI token limits
References