Session Archiver

Turn raw session transcripts into clean, searchable markdown archives.

When to use

• End of a long work session — distill key decisions, actions taken, and open items
• Preparing a daily memory file (memory/YYYY-MM-DD.md) from conversation history
• Archiving a completed project thread for future reference

What it does

1. Extracts key events, decisions, and action items from session text
2. Tags entries with categories (decision, action, blocker, insight)
3. Writes structured markdown to a target file (default: memory/YYYY-MM-DD.md)

Prerequisites

• read access to the session transcript or conversation text
• Write access to the target archive file
• The date command for timestamping

Steps

1. Read the source material

read path/to/session-log.md

If the session is this session, summarize from memory instead of re-reading.

2. Extract structured entries

Parse the text into categorized bullets:

• Decisions — choices made, with rationale
• Actions — things done, with outcome
• Blockers — what stopped progress, resolution status
• Insights — lessons learned, patterns noticed

3. Write to archive file

Format:

# YYYY-MM-DD Session Log

## Decisions
- [decision description] — rationale

## Actions
- [action description] — result

## Blockers
- [blocker description] — resolved/unresolved

## Insights
- [insight description]

Write to memory/YYYY-MM-DD.md (append if file exists).

4. Cross-reference

If a decision has a linked PR, file, or external URL, include it inline. Use absolute workspace paths for local files.

Notes

• Never archive secrets, tokens, or private credentials — redact first
• Keep entries terse; this is an index, not a transcript
• If the session is still active, note what's still in progress rather than finishing prematurely