Commit a conversation to working memory — extracts both project state and knowledge (insights, lessons, mental models, gotchas, recipes, corrections). Auto-fires when a user says /remember, "save this conversation", "commit this to memory", "log this session", or signals they're wrapping up. Also triggers if the user asks Claude to "remember" something specific about a project, decision, or piece of knowledge.
From session-memorynpx claudepluginhub brightwayai/session-memory-pluginThis skill uses the workspace's default tool permissions.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Searches prompts.chat for AI prompt templates by keyword or category, retrieves by ID with variable handling, and improves prompts via AI. Use for discovering or enhancing prompts.
Guides agent creation for Claude Code plugins with file templates, frontmatter specs (name, description, model), triggering examples, system prompts, and best practices.
See commands/remember.md for the full workflow.
When this skill fires automatically:
If the user said "remember that X" → determine if X is project state (decision, action) or knowledge (insight, gotcha, model). Commit to the appropriate entry type in the relevant node.
If the user is wrapping up → offer: "Want me to commit this session to memory?" If they agree, run full extraction covering both project state AND knowledge.
If mid-conversation → checkpoint: commit what's happened so far. Note
[CHECKPOINT] in the changelog.
If the conversation was purely exploratory/learning (no decisions or actions) → focus extraction on the Knowledge section. It's totally valid to commit a session that only produced INSIGHT/LESSON/MODEL entries with no project state.
If multiple projects → commit to each separately, flag cross-project signals.
If new node → note [NEW NODE] in changelog, write thorough initial summary.
Key extraction targets: