j-next

name	j-next
description	Prioritize open TODO items by effort and impact, recommend what to work on next. Add the `full` keyword to also weigh project notes and stash context.

/j-next

Recommend what to work on next from the open task list. By default this is fast: it reads only git and docs/TODO.org. The full keyword additionally weighs project notes and stash context to sharpen estimates and catch hidden blockers.

Arguments

/j-next [full] [focus]

• full — optional leading keyword. When present, after the fast ranking also scan docs/notes/ and (for shared-tooling focus) /workspaces/stash/notes to refine effort/impact and surface blockers the TODO list alone won't show (step 1b). Without it, rank from git + TODO only.
• [focus] — optional keyword(s) or short phrase to narrow scope (e.g., jolo skills, emacs, security, perf grafana). When given, rank only TODOs that match the focus or are direct prerequisites for that focused work. Anything after a leading full is treated as focus.

Without arguments, rank all open items from git + TODO.

Instructions

1. Gather context (always)

• Read docs/TODO.org. Treat as actionable: TODO, NEXT, and INPROGRESS. Treat BLOCKED and WAITING as not startable — keep them out of the ranking, but list them once afterward so the user sees what's parked and why.
• Run git log --oneline -n 10 to see recent momentum (what area was last worked on). The wider window than j-save/j-resume's -n 5 is deliberate: this read is for area detection, not a what-shipped recap.
• Note which items carry the :autonomous: tag — these are agent-runnable unattended (jolo autonomous).
• If a focus argument is given, split it into useful keywords and match against TODO titles, TODO body text, tags, referenced branch names, and recent commit subjects.

1b. Enrich from notes (only if full given)

Skip this entire step unless full was passed. Otherwise:

• Scan docs/notes/ for decision/research/gotcha notes that change an item's effort, impact, or reveal a blocker not recorded in the TODO. Read only the bodies of clearly relevant notes.
• If the focus (or the highest-ranked items) touch shared tooling or workflow (Emacs, org-mode, denote, devcontainers, agent behavior, host integration, or global perf services), also skim relevant stash note filenames in /workspaces/stash/notes.

emacsclient -e '(bergheim/agent-denote-list "docs/notes" 15)'

2. Assess each actionable item

For each item, estimate:

• Effort: small (< 1 hour), medium (1-4 hours), large (4+ hours)
• Impact: how much it improves the project
• Momentum: is it in the same area as recent work (lower context-switch cost)

Base effort estimates on what you can see in the codebase — check if referenced branches, files, or partial work already exist. An INPROGRESS item is usually the strongest candidate: finishing started work beats opening a new front.

If a focus argument was given, exclude unrelated items from the ranked table unless they block the focused work. Say "no open TODOs match this focus" when that is true, then recommend the nearest useful prerequisite if one exists.

3. Present the list

Print a ranked table, ordered from least effort to most effort. Mark autonomous-eligible items in the Auto column:

Effort   Auto   Item                                    Notes
──────   ────   ─────────────────────────────────────   ─────────────────────
small    ✓      Fix X                                   Branch exists, 1 file
medium          Add Y support                           Needs research
large           Rework Z                                Touches 5+ files

If any INPROGRESS items exist, list them first regardless of effort. After the table, list any BLOCKED/WAITING items in one line each with what they're waiting on.

After the table, recommend one item to start with and briefly explain why (effort/impact/momentum tradeoff). If a focus argument was given, make the recommendation within that scope.

4. Offer to start

Ask if the user wants to begin working on the recommended item. If the recommended item is autonomous-eligible, mention it can be dispatched unattended via jolo autonomous.

Rules

• Read-only: do not modify any files
• Be honest about effort — don't underestimate to make items look appealing
• If a TODO references a branch, check if it still exists before claiming partial work
• Never rank BLOCKED/WAITING items as startable
• If TODO.org is missing or empty, say so and suggest creating one