agentleFS
AN0099AN0099/toothpaste-kit/skills/daily-review/SKILL.md

daily-review

Skill1 starsChanged 20 days agoApache-2.0View on GitHub

Daily pass over the task board; empty the Inbox, check priorities, carry Doing forward, offer Verify for close, order Next and pick the Most Important Task. The person decides every move. Does not replace daily-dashboard, the state view.

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 10 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Priority is an Eisenhower quadrant: Do, Decide, Delegate, Delete.

    line 33
  2. Distinctive:

    Run the lane-board command once and read all of it, including the coverage line.

    line 40
  3. Distinctive:

    Verify, and that is a finding, not a close.

    line 59
  4. Distinctive:

    Quadrant check: Every open task (Inbox, Next, Waiting, Doing) without a quadrant gets a proposed one.

    line 60
  5. Distinctive:

    Next, ordered: Do first, then Decide, in the board's order.

    line 63
  6. Distinctive:

    Next actions: Every task in Next, Waiting or Doing carries one concrete next action, since a task moves between lanes by that action.

    line 68
  7. Distinctive:

    Keep each table to the tasks the step moves.

    line 72
  8. Distinctive:

    Write the accepted order of Next as each task's rank, 1 upward, and clear the rank of a task leaving Next.

    line 84

“Common” is measured: kept by at least 6 in 10 of the 10 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

When to Use

Once a day, before the first task is pulled from Next. Invoked by a person, never by an agent.

Adapting this skill: filenames below name roles rather than literal paths. It assumes a task queue with lanes (below) and a lane-board tool: one command that prints every task grouped by lane, with its priority, rank, assignee, carry count, due date and consequence, blocker and next action, orders Next by quadrant and then rank, validates the queue before printing, and ends with a coverage line naming the fields it did not read. Check that line before starting: a field this review needs and the tool does not print is read by an ad hoc command, which is the failure the tool exists to prevent, so extend the tool instead. CLAUDE.md names the command. A tree without such a tool builds it before running this skill, for the reason orient gives: a model re-deriving lanes from raw JSON is doing a script's job, and doing it inconsistently.

The lanes

LaneHoldsLeaves by
Inboxcaptured, not yet clarifiedthis review
Someday/Maybeconsidered and parkedmonthly-review
Nextready to start, in ordera seat pulling its top task
Waitingblocked on a named person, task or eventthe blocker clearing
Doingpulled; at most one per seatfinishing, to Verify
Verifydeliverable exists and a second party checked itthe person's word, to Done
Done, Archiveaccepted; out of the boardmonthly-review archives

Priority is an Eisenhower quadrant: Do, Decide, Delegate, Delete. Urgent is read, not felt: a task is urgent only when it carries a due date and a stated consequence of missing it. Important means it serves the long-term goals the tree has written down; where none are written, importance is the person's call, and the review says so rather than inventing a basis.

Phase 1: Observe

Run the lane-board command once and read all of it, including the coverage line. If it fails validation, stop and report the fault: a board drawn from an invalid queue is a claim about the queue, not the queue.

Phase 2: Propose, lane by lane

For each step, a short table of proposed moves, one row per task, each with a one-line reason. Do not write anything yet.

  1. Inbox to empty. Each item gets a destination: Next (with a quadrant and a next action), Waiting (with its blocker), Someday/Maybe, a child of an existing task, or trash. Work that takes under two minutes is proposed as done now and noted in the capture file, not queued.
  2. Doing, carried. Each task still in Doing from a previous day: carry it (it stays, and its carry count rises by one), move it back to Next, or move it to Waiting with a blocker. A task bumped by an interruption returns to the top of Next and its count does not rise. A count of three or more is said aloud: a task carried all week is a question for a weekly review, and the daily review names it so the question is not first asked there.
  3. Verify, offered for close. Each task in Verify, with the second-party check its note records. Offer the close; never close it. A task whose note records no second party does not belong in Verify, and that is a finding, not a close.
  4. Quadrant check. Every open task (Inbox, Next, Waiting, Doing) without a quadrant gets a proposed one. A task marked urgent with no due date or no consequence gets a proposal to drop to Decide, in whichever open lane it sits: a Waiting task keeps its quadrant when it returns to Next.
  5. Next, ordered. Do first, then Decide, in the board's order. The Most Important Task is the first task in Next for the person's own seat, in the order this step proposes, which the board shows once phase 4 writes the ranks; the board marks it in words. Name it with its next action. For each agent seat whose Doing is empty, name its first pull, or say that Next holds nothing for that seat.
  6. Next actions. Every task in Next, Waiting or Doing carries one concrete next action, since a task moves between lanes by that action. Propose one for each task the board shows without it. For a Waiting task, the action is whatever clears the blocker, and it names who takes it.

Keep each table to the tasks the step moves. A step with nothing to propose is one line saying so.

Phase 3: Ask, then stop

Chat output, not a document: the dashboard's phase 0 applies here too. End with questions. One per step that has proposals, each accepting that step's table as written; the person answers exceptions by task ID in prose. Then stop.

Phase 4: After the answers

Apply exactly the accepted moves, in one scripted edit that asserts each task's current lane before changing it, so a task that moved since the board was read fails loudly instead of moving twice. Write the accepted order of Next as each task's rank, 1 upward, and clear the rank of a task leaving Next. Record each move in the task's note with the date and "daily review". Run the lane-board command again and show the lanes that changed. Nothing moves on a task the person did not answer; an unanswered row stays where it was and is listed as unanswered.

Scope Pointer

  • daily-dashboard: the state view. It checks the standing files and working trees against their claims; this reads the board and moves tasks. Run it first when both are wanted.
  • weekly-review: Waiting's blockers, carry-over by count, the Done lane, and one Decide task scheduled so the important is not starved by the urgent.
  • monthly-review: Someday adopted or rejected, and archive candidates chosen by the person.
  • session-close: updates the standing files at the end of a session. A move made here is already in the queue and does not wait for it.
<!-- SPDX-FileCopyrightText: 2026 Aidan Naveja SPDX-License-Identifier: Apache-2.0 -->

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.