Session Cleanup

Overview

Use this skill to identify and archive old, low-signal sessions based on user-message count and age, while keeping any session with meaningful content or metadata.

Workflow

1. Decide the cutoff (days ago or explicit date) and user-message threshold.
2. Run the cleanup script in dry-run mode to review candidates.
3. Re-run with --apply to archive sessions that are still low-signal.

Heuristics

Treat a session as low-signal if ALL are true:

• user_message_count <= max_user_messages (after removing boilerplate)
• Session is older than the cutoff date
• No name, tags, or notes
• No meaningful user/assistant text (beyond boilerplate)

Boilerplate includes: AGENTS.md instructions, environment context blocks, empty strings, and single "." messages.

Quick Start

Dry-run for sessions older than 1 day with at most 1 user message:

bash
1python3 /Users/innoiso/.codex/skills/session-cleanup/scripts/cleanup_sessions.py --min-age-days 1 --max-user-messages 1

Apply archiving with the same filters:

bash
1python3 /Users/innoiso/.codex/skills/session-cleanup/scripts/cleanup_sessions.py --min-age-days 1 --max-user-messages 1 --apply

Target an explicit cutoff date (UTC):

bash
1python3 /Users/innoiso/.codex/skills/session-cleanup/scripts/cleanup_sessions.py --cutoff-date 2026-01-24 --max-user-messages 2

Output

The script prints:

• total candidates
• kept vs archived
• sample IDs for review (in dry-run)

Resources

scripts/

• cleanup_sessions.py: paginate sessions, count user messages, detect boilerplate, and archive eligible sessions.