Injections
Injections are shared fragments inserted into prompts via components. Update the
YAML once — all prompts that reference it receive the updated rules automatically.
All injection files live in prompts/injections/.
Full index
| Injection | Type | Used in | Source doc |
|---|---|---|---|
structure_instructions | Static | Structure Steps 3, 4b, 5 | Course Design Standards |
naming_instructions | Static | Structure Steps 3, 4b, 5 | Naming Conventions |
assessments_instructions | Static | Structure Steps 3, 4b, 5 | Course Design Standards |
how_to_pick_random_name | Static | Structure Step 1 | — |
persona_model | Static | Audience analysis, Structure Step 1 | Persona model |
language_instructions | Dynamic | Section content: text, edit | — |
avoid_ai_language | Static | via language_instructions | — |
html_instructions | Static | Section content: edit | — |
artifact_framework | Static | Section content: text | — |
video_backgrounds | Static | Section content: text | — |
image_brand | Static | Image: illustration, diagram, checker | Image Pipeline |
Dynamic injections
Most injections are static text. One exception:
language_instructions has its own variables (language, localization) and
loads avoid_ai_language as a component. It combines prose quality rules with
target-language instructions (CEFR B1, acronyms, forms of address).
Where to find details
- Structure pipeline injections (
structure_instructions,naming_instructions,assessments_instructions): rules are documented in Course Design Standards and Naming Conventions. - Section content injections (
language_instructions,html_instructions,artifact_framework,video_backgrounds): usage is documented in Section Content Pipeline. persona_model: documented in Persona model.how_to_pick_random_name: 7-step sampling algorithm for random European names. No separate doc — logic is self-contained in the YAML.