context-looplisted

# Context Loop A session continuity protocol that uses AGENTS.md as Claude's persistent memory across conversations. Claude has no memory between sessions — this skill is the fix. The loop has two moments: **session start** (restore) and **session end** (save). Both are required. Skipping the end means the next session starts blind. Naveen works in two modes. The skill behaves slightly differently in each. --- ## Mode Detection — Chat vs Project Before applying the loop, Claude should detect which mode is active: **Chat Mode** — a standalone conversation, not inside a Claude Project. Naveen typically converts interesting chats into Projects once they reach a meaningful conclusion or produce something worth continuing. **Project Mode** — inside a Claude Project. Project Knowledge is available. AGENTS.md lives in Project Knowledge, not on Naveen's desktop. Static files (skills, preferences, shortcodes) are already loaded automatically. If unsure which mode is active, check whether Project Knowledge is present in context. If it is — Project Mode. If not — Chat Mode. --- ## Chat Mode Used for standalone conversations. Naveen pastes AGENTS.md manually each session. ``` SESSION START └── Naveen pastes AGENTS.md into chat └── Claude reads it, confirms understanding, begins work ... session happens ... SESSION END └── Naveen types: /learnings └── Claude writes structured update (see format below) └── Naveen copies the output └── Naveen pastes it into AGENTS