📖About & Usage

About

Carol Context is a private, administrator-only inspection app for seeing what Carol had available during a particular session. Think of it as an “open the suitcase and check everything inside” view: it shows the session induction—the starting information that orients Carol—the system prompt, which sets her core instructions, the tools she could use, and any reused context kept through caching. It exists so administrators can understand Carol’s working environment without relying on guesswork.

Owned by Carol, Carol Context complements operational views such as Carol Monitor and the introductory material managed through Session Induction. Its focus is not what Carol ultimately said or did, but the information and capabilities loaded before and during that work. This makes it useful for accountability, troubleshooting, and checking whether a session began with the intended instructions.

Usage Patterns

Carol Context is used when an administrator needs a full context audit of one session. It may be opened after an unexpected answer, a tool-use problem, or a change in behaviour to check whether the right induction, instructions, and tools were present. It can also reveal whether cached material—previously prepared context reused for efficiency—affected what Carol saw.

For example, suppose Carol gives an answer that appears to overlook a current policy. An administrator can inspect the relevant session in Carol Context and ask practical questions: Was the expected instruction loaded? Did the session receive the correct tool list? Was older cached context reused? The app does not itself judge the answer; it provides the evidence needed to investigate it carefully.

🗂️Tabs & Screens

Tab inventory is being built — see CAROL-INI-077 step 7.

👤Owner

📚Recent initiatives

Initiatives that touched this app — a short summary each; open one for the full story.