The Accountability Kit.

In Loop & Gate, this kit points the whole method at you. A daily loop can remind, track, and score on its own, and it should. But it can't want the outcome for you, it can't decide a task is worth killing, and it can't tell you a streak is theater. So the loop stops at a fixed set of points and waits for you. Those points are the gates.

The engine underneath does the one mechanical thing willpower forgets: every morning focus that doesn't show up in the day's log gets a deferral increment, and at three strikes it trips a pattern alert that forces a decision. That mechanic, not the note-taking, is the whole point.

It is not a habit tracker that rewards streaks, and it is not an autopilot: it surfaces the reckoning and counts the misses, but the follow-through is yours. Answer it with a comfortable story and all you get is a well-documented record of your own avoidance.

Nothing else to install. No pipeline, no account, no local model, no subscription beyond your Claude plan. The reckoning, the counting, and the weekly aggregation are done by the agent you're already talking to, over plain text files.

Each gate names its mode — Commit for deciding what deserves you, Reckon for facing what you did, or Pattern for reading the trend — so you know what kind of call you're making. Gate ∞ decides how much of the rest today earns: a day with one exhausted hour earns a one-line log.

You don't run all of it every day. On a brutal day, /log one line and move on. On a normal day, morning and night. On a Sunday, the weekly pass.

Six steps, and only one of them is bookkeeping:

1. Morning names the target. Gate 1 writes one primary focus, and up to two secondaries, to today's daily note.
2. The day gets captured. Gate 2 appends one-line log entries. Cheap on purpose — the log is the evidence the reckoning reads.
3. Night reckons and counts. Gate 3 diffs the focus against the log and increments a deferral counter for every focus item the log doesn't show as done.
4. Three strikes trips an alert. At a count of 3 the reckoning stops being polite and forces a choice: re-commit with a reason and a slot, or cut it.
5. Named patterns get checked. The avoidance patterns you wrote in goals.md — “infrastructure over output,” say — are checked at the morning and weekly gates, and flagged by name when they match.
6. The week is scored straight. Gate 6 aggregates tasks deferred twice or more, days with no log at all, and the share of logged work that maps to a stated goal. A low alignment score is the mirror, not an error.

The whole engine converts a vague background guilt into a specific number and a forced choice. The deferral log itself is one row per tracked task in patterns.md, with a count, the last date, a category and a note — struck-through rows are the completion log, kept for history.

The source lists the asset as complete: the gates reference, the deferral-engine mechanic, the five ritual skills, the templates, and the no-terminal on-ramp. It runs on Claude alone, and the Obsidian-vault path is auto-detected when one is present.

Where state lives is deliberately disposable — plain markdown in the project folder by default, your vault when you keep one. The gates are the part meant to last: when the files get rebuilt, they don't change.

The kit stands alone, but if the Build and Grow kits are installed, the nightly reckoning is where a stalled build or an unposted launch gets caught — which is how Build → Grow → Accountability → Build closes.