jouet wind review

• Usage: jouet wind review [FLAGS] [SESSION]

Digest whole sessions: corrections, failures, waits, re-reads

Reads a Claude Code or Codex transcript and reduces it to a few thousand tokens: the user's corrections and interruptions, failures and their retries, long calls, where tokens and time went (as jouet audit counts them), files read repeatedly, commands typed out again, and a timeline with a phase per prompt. Every item names its lines in the transcript. Secrets are redacted (with gitleaks' rules too, when a config is found) and the home directory is written as ~. Nothing leaves the machine. jouet wind review prints the same with a prompt for a reviewing agent.

Arguments

• [SESSION] — A session ID, a prefix of one, or a transcript's path

Flags

• --last — The most recently written session (inside a session, that session)
• --since <TIME> — Every session written since TIME (7d, 12h, a date or RFC 3339), the 8 most recent
• --harness <HARNESS> — Which transcripts to read: claude, codex or all (default all)
• --project <PATH> — Only sessions that ran in PATH or a directory below it
• --budget <N> — Tokens per session's digest, about (default 4000)
• --lease <LEASE> — Use the recording instructions for a claimed scheduled review
• -h --help — Print help