name: save description: Quick-save key findings, decisions, or notes to the sayou workspace user_invocable: true triggers:

• "save this"
• "remember this"
• "save findings"
• "save decision"
• "save note"
• "save what we learned"

/save — Quick Save to Workspace

Save key knowledge from the current conversation to the sayou workspace as a structured, versioned file.

Instructions

1. Analyze the conversation — identify the most important knowledge to save:

   • Decisions made (and their reasoning)
   • Research findings
   • Bug fixes and root causes
   • Architecture choices
   • Meeting notes or action items

2. Choose a logical path based on content type:

   • decisions/ — architectural decisions, technology choices, design rationale
   • research/ — competitive analysis, market research, technical investigation
   • notes/ — meeting notes, brainstorms, general notes
   • bugs/ — bug investigations, root causes, fixes applied
   • guides/ — how-to guides, setup instructions, runbooks

3. Write with YAML frontmatter — always include structured metadata:

   ---
   type: decision | research | note | bug-report | guide
   status: draft | reviewed | approved | archived
   topic: short-topic-name
   date: YYYY-MM-DD
   tags: [relevant, tags]
   ---
   

4. Use workspace_write to save the file. Use a descriptive filename in kebab-case (e.g., auth-token-refresh-fix.md).

5. Confirm what was saved — show the path, key frontmatter fields, and mention it's versioned (they can update it later and the old version is preserved).

Example

If the conversation discussed choosing PostgreSQL over MySQL:

---
type: decision
status: approved
topic: database-selection
date: 2026-02-19
tags: [database, infrastructure]
---
# Database Selection: PostgreSQL

## Decision
PostgreSQL over MySQL for the new service.

## Reasoning
- Better JSON support (JSONB) for our flexible schema needs
- Superior full-text search (eliminates need for Elasticsearch)
- Row-level security for multi-tenant isolation

## Alternatives Considered
- MySQL: Familiar to team but weaker JSON/FTS support
- MongoDB: Overkill for our relational data patterns

Save to: decisions/database-selection.md

Tips

• If the user says "/save" without context, scan the conversation for the most significant knowledge and propose what to save
• If multiple things are worth saving, save them as separate files
• Always use workspace_write (not regular Write tool) — this goes to the persistent workspace, not the local filesystem
• If a file at the chosen path already exists, workspace_write creates a new version (nothing is overwritten)