journal

name

journal

description

Use for any claude-journal pipeline command. The user runs `/journal <action>` where `<action>` is `setup` (first-time data-repo bootstrap), `schedule` (create or update the nightly consolidator routine), `consolidate` (distill recent breadcrumbs now, on demand, including the current session, instead of waiting for the nightly run), `recall` (answer a question from the journal's digests/memories, e.g. "what did I do yesterday" or "what do I know about project X"), or `accept` / `skip` / `edit` (resolve pending proposals surfaced at SessionStart). Also triggers when the user asks to set up, bootstrap, or schedule the journal / consolidation pipeline, to consolidate or distill recent work now, to recall or ask what they did on a day or what the journal knows about a topic, or to accept, skip, or edit its proposals.

You are handling a /journal <action> invocation. The first word of the invocation argument (the $ARGUMENTS slash-command placeholder) selects the subcommand; everything after it is that subcommand's own argument. Each flow lives in its own reference file next to this one — read only the one for the chosen action, then follow it as if it were the whole skill.

<action>	What it does	Then
setup	One-time, first-machine bootstrap of the claude-journal data repo (tools, git identity, gh sign-in, the GitHub token the cloud routine needs, then the bootstrap).	Read references/setup.md (relative to this skill's directory) and follow it exactly.
schedule	Create — or update — the once-per-account nightly consolidator routine via /schedule.	Read references/schedule.md and follow it exactly.
consolidate	Distill recent breadcrumbs now, locally, including the current session, instead of waiting for the nightly routine. Runs the same Tracks 0–3 against the auto-detected un-consolidated dates (or one explicit YYYY-MM-DD), then commits and pushes.	Read references/consolidate.md and follow it exactly.
recall	Answer a question from the journal: time questions (what did I do yesterday) from digests/, topic questions (what do I know about X) from memories/. Reads distilled outputs only, never raw/; offers /journal consolidate for any day not yet distilled.	Read references/recall.md and follow it exactly.
accept / skip / edit	Resolve the pending proposals (new skills, feedback memories, CLAUDE.md edits) for the current project; pass the action through.	Read references/resolve.md and follow it exactly.

Rules:

• The five references are mutually exclusive flows. Don't read more than one, and don't merge their steps.

• setup and schedule are consequential one-time actions. Their references carry their own confirmation gates, idempotency checks, and key/token guardrails — honor them; don't shortcut.

• If there is no action (bare /journal) or an unrecognized one, don't guess — print this help and stop:

  /journal setup              bootstrap the data repo (run once, first machine)
  /journal schedule           create or update the nightly consolidator routine
  /journal consolidate [date] distill recent breadcrumbs now (incl. this session)
  /journal recall <question>  answer from the journal (e.g. what did I do yesterday)
  /journal accept|skip|edit   resolve pending proposals for this project