Problem
AI context is invisible. Can't tell what rules AI is following, whether it read your ground rules, or if context has degraded.
Pattern
Use visual markers (emojis) to signal active context:
- Start every response with a marker showing current mode
- Different markers for different contexts/roles
- Stack markers when multiple contexts active
- Special markers for specific actions (errors, re-reads)
- Can be impromptu one-offs for crucial instructions (when adding an important instruction mid-conversation, ask it to reply to you with an additional emoji)
Makes the invisible parts of context visible at a glance.
Example
Some example markers:
- 🍀 = ground rules have been read
- 🔴/🌱/🌀 = Shows specific TDD (red|green|refactor) phase and that it read tdd.md process file
- ✅ = committer role active
- ❗️ = flagging an error
- ♻️ = rules just re-read
- ✨📂 = creating new repository
How to set this up
In ground rules (user level):
**ALWAYS** start replies with STARTER_CHARACTER + space (default: 🍀). Stack emojis when requested, don't replace.
In specialized contexts (committer):
When I tell you're a committer, add ✅ to STARTER_CHARACTER emojis. Make sure there's a space between any emojis and the text
In process files (TDD):
STARTER_CHARACTER = 🔴 for red test, 🌱 for green, 🌀 when refactoring, always followed by a space
This way let them stack: "🍀 ✅" = base rules loaded + committer role active. Easy to see what context AI is operating under.
