CCE User Guide
Context and Cognitive Engineering for prompts, context packs, memory, profiles, test suites, token analysis, regressions, and A/B tests.
Who This Guide Is For
- Prompt engineers
- AI product teams
- Evaluation teams
Where To Go
| Page |
Use It For |
/cce |
CCE dashboard. |
/cce/builder |
Build context and prompt assets. |
/cce/playground |
Test prompts and context. |
/cce/context-packs |
Manage reusable context packs. |
/cce/memory |
Manage memory assets. |
/cce/templates |
Prompt and context templates. |
/cce/test-suites |
Regression and quality test suites. |
/cce/prompt-comparisons |
Compare prompt versions. |
/cce/ab-tests |
Run A/B tests. |
/cce/tokens |
Analyze token usage. |
Core Concepts
| Concept |
Meaning |
| Context pack |
A reusable set of instructions, facts, examples, and constraints. |
| Profile |
A target persona, behavior, domain, or operating mode. |
| Prompt comparison |
A side-by-side test of prompt/context variants. |
| Regression |
A previously passing scenario that fails after a change. |
| Token budget |
The size and cost constraint for prompt, context, and response. |
Common Workflows
Create a production prompt asset
- Create or select a template.
- Attach context packs and profile.
- Test in the playground.
- Run prompt comparisons.
- Run test suites.
- Review token budget.
- Publish only after regression checks pass.
Investigate regressions
- Open Regressions.
- Find the failing scenario.
- Compare the last passing and current context.
- Use Context Diff to identify changes.
- Patch prompt or context pack.
- Rerun the test suite.
Best Practices
- Version prompts and context packs together.
- Use test suites for critical user journeys.
- Watch token budgets before adding large context blocks.
- Use A/B tests for behavior changes that affect users.