Tracking tickets in git

Your sprint state lives in .tickets/: each ticket's acceptance criteria, plan, evaluator report and summary. In a project that uses canon, commit it and do not add it to .gitignore. It is the durable record of each sprint, and it survives a context reset.

Why it must be committed

The dialog Commit the ticket before creating the worktree, listing the ticket's own files as required and ticked, other tickets' files as uncommitted and not included, and the commit message chore: add ticket t-tko0.
The Cockpit commits a ticket's own files before it creates a worktree.

If you started from a template that gitignores .tickets/, remove that line so your agent's plans and specs travel with the repo.

Do not commit the runtime files

canon also writes a few per-machine files under .tickets/ that change every session:

tkt seeds a .tickets/.gitignore covering them, with the two lines .cockpit-* and ACTIVE. It does this when it sets up the tickets folder and the file is absent, and never overwrites a .gitignore you already have. Commit that file, so the ticket docs are committed and the runtime churn is ignored.

If your repository already committed any runtime files, from before the .gitignore existed, untrack them once. They stay on disk and only stop being versioned:

git rm -r --cached --ignore-unmatch '.tickets/**/.cockpit-*' '.tickets/ACTIVE' git commit -m "stop tracking canon runtime files"

The session log is optional

Each ticket can also hold a cockpit-sessions.md, a one-line log of where each session ran, for example "2026-10-03: sprint start used main checkout". It is not covered by the seeded .gitignore. canon's notes on the worktree dialog above call it optional: "session logs like cockpit-sessions.md are optional", while the ticket's own docs are required. Commit it if you want the history, and leave it untracked if you do not. Either way works.