Imported from streamient/streamient (
SKILL.md). Install upstream withnpx skills add streamient/streamient. Copyright stays with the author.
Streamient Memory
Use Streamient as a private, cross-session context layer. Retrieve only relevant context and save only durable outcomes.
Retrieve context
- Form a focused query from the task.
- Use
recall_memoryfor prior decisions, preferences, debugging insights, and task outcomes. - Use
search_notesfor specifications, documentation, meeting notes, and longer structured material. - Start with three results. Broaden the query or increase the result count only when results are weak.
- Use
read_memoryorread_notewhen the complete selected item is needed. - Use
list_projectsorget_projectonly when project scope is unclear. Never guess a project ID. - Apply retrieved context only when relevant. Prefer newer explicit user instructions when they conflict with stored context.
Preserve outcomes
Write only when the task produces reusable context:
- Use
store_memoryfor concise decisions, preferences, debugging solutions, conventions, and task outcomes. - Use
create_notefor structured specifications, architecture decisions, plans, and meeting notes. - Call
suggest_memory_tagsbeforestore_memory; reuse existing descriptive tags when suitable. - Include enough source or task context to make the stored item understandable later.
- Skip writes for trivial answers, transient status, raw logs, or information already stored.
Safety and quality
- Do not store passwords, access tokens, API keys, authentication codes, or other credentials.
- Do not store sensitive personal information unless the user explicitly requests it and confirms the intended content.
- Keep retrieval and writes within the user's authenticated Streamient workspace and selected project.
- Do not claim that context was retrieved or stored unless the corresponding tool call succeeded.
- If Streamient is unavailable, continue when possible and state briefly that persistent context could not be accessed or saved.