扫码联系在线客服Reference Session
Import context from past sessions into current work without duplicating full context.
Input: $ARGUMENTS
Process
Step 1: Parse Arguments
Parse for: search term (session name fragment, Linear issue ID, or topic keyword), optional content type (decisions, context, or all; default: decisions).
/reference-session auth # Search "auth", import decisions
/reference-session auth context # Import full context summary
/reference-session PLT-123 # Search by Linear issue
/reference-session # List available session decisions
Step 2: Search for Sessions
Option A: Use Serena MCP list_memories(), filter for session-*-decisions.md
Option B: Search .agents/sessions/archive/*<search-term>*.md
Step 3: Display Matches
Show matching sessions with date, Linear issue, and decision count. If single match, proceed directly.
Step 4: Read and Import
Read decision memory via Serena or extract from archived session file.
If active session exists: update session frontmatter referenced_sessions and add ## Referenced Decisions section with decision name, what was decided, rationale, and context.
If no active session: display decisions without updating any file.
Step 5: Report
Confirm what was imported: source session, content type, decision count, summary list, and updated session path.
Content Types
| Type | Imports |
|------|---------|
| decisions (default) | Decision name, what was decided, rationale, council outcome |
| context | Problem statement, technical approach, key files, final outcome |
| all | Both decisions and context |
Anti-Patterns
| Don't | Do Instead |
|-------|------------|
| Import full session files | Import only decisions/summaries |
| Import without tracking | Always update referenced_sessions |
| Import from unrelated sessions | Search by topic or Linear issue |
Related Skills
- resume-session -- Resume active sessions (not reference past ones)
- review-sessions -- Session hygiene and cleanup
- implement -- Start new work that may reference past sessions