recall-contextlisted

# Recall Context Proactive memory retrieval that pulls prior context relevant to the work about to happen. Reduces "starting from scratch" errors and prevents the agent from re-deriving conclusions already captured in mnemo. ## When to Use - **Session start**: load any preferences, decisions, or open questions tied to the current cwd / project before the user types their first prompt. - **Before a significant decision**: surface prior decisions in the same area (e.g. database choice, lint rules, deployment target) so the agent does not contradict an earlier conclusion. - **When the user names a topic**: e.g. "let's work on the auth flow" - pull memories tagged or describing auth before proposing a plan. - **After a long context gap**: if the conversation referenced earlier decisions but the agent does not have them in working memory, recall them on demand. ## Steps 1. **Resolve query terms** from the trigger: - `cwd`: use the current working directory path + project name as the query (e.g. `mnemo-mcp` or `/c/Users/.../wet-mcp`). - `recent`: use the last 5-10 file paths the agent edited or read. - `<topic>`: use the topic verbatim (the user's words). - Default (no arg): combine cwd + last 3 file paths. 2. **Search mnemo** with `context_type` filtering when applicable: ``` memory(action="search", query="<resolved query>", context_type=null, limit=10, include_archived=false) ``` - For decisi