agentleFS
jwm-axonijwm-axoni/claude-workbench/skills/weekly-review/SKILL.md

weekly-review

Skill2 starsChanged 7 days agoMITView on GitHub

Workbench routine. On the first workday of the week it writes the week-ahead plan to Daily/YYYY-MM-DD_NEXT_WEEK_PLAN.md from last week's recaps, runs the librarian's checks and fixes what they find (frontmatter, links, index rows, stale claims, unfiled captures), makes every move into Archive/, grades the dream, and tunes the confidence bar on evidence in Tracking/Signals/YYYY-MM-DD_weekly.md. Runs as "[Workbench] Weekly Plan & Maintenance"; can be run by hand with /workbench:weekly-review.

92 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use quarterly-review, daily-plan and 8 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Phoenixrr2113Skill

    quarterly-review

    Quarterly review — strategic look at the last 3 months. Career trajectory, project health, life balance, big decisions ahead. Runs 1st of Jan/Apr/Jul/Oct.

    Also in 3 projects that use this

  • steptianSkill

    daily-plan

    每日待办计划与日程管理。触发词:'今天要做'、'今日计划'、'今日待办'、'今天安排'、'安排一下'、'加个任务'、'记个任务'、'今日复盘'、'今日回顾'、'今天总结'、'本周计划'、'周总结'、'开始工作'、'上班了'、'完成了'、'加个日程'、'记个日程'、'改一下时间'、'日程调整'、'面试'、'开会'、'约了'。支持四象限优先级、时段分配、固定日程管理、智能时间估算、每日/每周复盘。当用户提到有明确时间的安排(如'下午两点面试'、'10点开会'),自动识别为日程并写入日程表。

    Also in 3 projects that use this

  • seedprodSkill

    daily-brief

    Send a morning briefing via Telegram with weather, news, and a motivational quote. Designed to run via launchd for automated daily delivery.

    Also in 5 projects that use this

  • claude-office-skillsSkill

    meeting-notes

    >

    Also in 8 projects that use this

  • Phoenixrr2113Skill

    morning-brief

    Full morning brief — calendar, email, meeting prep, execute actionable items, create follow-up tasks. Runs at 8:03 AM or when Randy asks for a morning update.

    Also in 4 projects that use this

  • zephyrwang6Skill

    daily-review

    每日工作回顾与洞察分析,对用户前一天的对话历史和使用数据进行总结与建议

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 0 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 80 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Run the checks from the librarian skill across every live folder, then fix what is safe to fix and list the rest in the plan's Health …

    line 71
  2. Very specific:

    Write Tracking/Signals/<today>_weekly.md with a Thresholds in force table (kind | threshold | precision | n | changed from | why).

    line 109
  3. Very specific:

    Put a Learning table in the Signals file and its headline in the plan's What the system learned.

    line 118
  4. Very specific:

    Log: every new item under Proposals in last week's DREAM_LOG entries, plus your own from FRICTION_LOG ## Open, as proposed with an id …

    line 138
  5. Very specific:

    Never close an open or looks-stale row.

    line 179

“Common” is measured: kept by at least 6 in 10 of the 80 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

Weekly plan and maintenance

Two jobs: tell the owner what the week ahead needs, and keep the vault healthy. You run when nothing else does, which is why you're the one routine allowed to move files and fix structure.

You own: Daily/<today>_NEXT_WEEK_PLAN.md, Tracking/Signals/<today>_weekly.md, Tracking/IMPROVEMENTS.md, frontmatter (type, tags, reviewed), link repairs, missing standard headings, Projects/INDEX.md row repairs, stale claims in page prose outside dream-owned blocks, every move into Archive/, and closing rows in Tracking/COMMITMENTS.md: nightly-dream adds rows and updates Due, Who owes, Last evidence and State; weekly-review alone removes rows, closing looks-done rows on evidence. You also append your one line to Tracking/Signals/RUN_LEDGER.md. You never write inside a (dream) block or a ledger the dream owns, any CLAUDE.md, another routine's Daily/ file, Captures/, or the contents of Archive/. You never delete.

Inputs

Root CLAUDE.md, Context/me.md (role lens), Tracking/GOALS.md, last week's Daily/*_DAILY_RECAP.md files, last week's Tracking/Signals/*_recap.md and *_dream.md, the last five DREAM_LOG.md entries, Projects/INDEX.md and active project pages, this week's calendar (calendar files or connector). Build on the recaps; don't re-sweep the sources.

1. The plan

---
type: weekly-plan
date: YYYY-MM-DD
tags: [weekly, plan]
status: active
source: claude-routine
---

# Week ahead — YYYY-MM-DD (week of … to …)

## Health
Routines that ran vs. expected last week (from `Tracking/Signals/RUN_LEDGER.md`, falling back to
Signals and Daily files for weeks before the ledger existed), captures unfiled for more than two
workdays, pages past their review date, open FRICTION_LOG items. Three to six lines.

## Priorities this week
Three to five, each tied to a goal in GOALS.md or a hard date.

## Carryover
Open commitments owed by the owner, oldest first, from project ledgers.
Then what the owner is waiting on from others, and who to chase.

## Role lens
The weekly checks listed under "Role lens" in Context/me.md.

## Meetings needing prep
## What the system learned
The scorecard headline (is it getting better?), rules the dream promoted last week, changes to the
confidence bar and why, and improvements done or verified.
## Proposals waiting on you
At most two, each one line with its evidence and a yes/no question.
## Needs reconciling
At most three, numbered, from the contradiction queue (step 7).
## Commitments closed
Rows closed this week (step 8), one line each with the evidence.
## Notes

2. Maintenance (the librarian's checks, fixed)

Run the checks from the librarian skill across every live folder, then fix what is safe to fix and list the rest in the plan's Health section:

FindingFix
missing type/tags/statusadd them from the folder's CLAUDE.md schema
path-style or broken wikilinksrewrite to bare form; repoint to the right page if unambiguous
a standard heading missing (e.g. ## Open commitments (mine) on a project)add it, empty
INDEX.md row out of step with a project pagefix the row to match the page
stale claim in page prose outside dream blocksdemote it to superseded with a forward link
page confirmed currentset reviewed: to today
capture lines unfiled > 2 workdayslist them in Health; the dream files them
near-duplicate pageslist them, propose a survivor; never merge unattended
one-sided relationship edge (project lists Pat, Pat doesn't list the project)add the missing inverse
## Related disagrees with the relationship propertiesrewrite the body lines from the properties
relationship property pointing at a page that doesn't existmove it to plain text in ## Related; list it for the dream
YAML wikilink not quoted (org: [[Acme]])quote it (org: "[[Acme]]"); unquoted breaks the YAML
dormant page (librarian check 30)list at most three in Health as yes/no questions; never change status or move it unattended

3. Archive moves

You're the only writer of moves into Archive/. Move, never delete, and mirror the folder:

  • a superseded ledger entry unreferenced for 30 days → cut it into Archive/Superseded/<page> ledger.md (append), leaving the forward links intact
  • a shipped project with no mention in 30 days → Archive/Projects/, and move its row to ## Archived in INDEX.md
  • a former organization quiet for 45 days → Archive/Organizations/, status: archived
  • a former person quiet for 90 days → Archive/People/

Log every move under ## Moved in your Signals file: what, from, to, when, which rule fired.

4. Measure and tune

From last week's *_recap.md (graded briefing claims) and *_dream.md (dream claims, graded against whether the owner kept, edited or reverted them in the pages and git history if available), compute per kind (status, commitment, priority, decision, person, term): kept, edited, deleted, unaddressed, and precision = kept ÷ (kept + edited + deleted).

Write Tracking/Signals/<today>_weekly.md with a Thresholds in force table (kind | threshold | precision | n | changed from | why). Move a threshold by at most 0.05 a week, and only with at least 5 graded claims of that kind: up if precision < 80%, down if precision ≥ 95% and many items sat in Needs your yes unanswered. If most claims are unaddressed (the owner isn't giving feedback), hold every threshold and say so; silence is not agreement. Anything that would leave the vault (send, post, change another tool) always asks, at any confidence.

5. Is it learning? The scorecard

Put a Learning table in the Signals file and its headline in the plan's What the system learned. Compare this week with the previous four from earlier *_weekly.md files:

MeasureFromGood direction
Draft acceptance: % verbatim + light-edit of graded draftsrecap ## Draftsup
Heavy edits per weekrecap ## Draftsdown
Briefing precision by kindrecap claimsup, or steady above 85%
Repeat corrections: a correction on a topic that already has a learned rulecaptures, voice/me rulesdown (a repeat means the rule isn't working: mark it not-working and say why)
Questions answered: % of digest questions with an answer within 2 workdaysrecap ## Answers, answer: capturesup; if low, ask fewer
Proposals adopted / verifiedIMPROVEMENTS.mdup

Say in one or two plain sentences whether the workbench is getting better and what's holding it back. If there's too little data (fewer than 5 graded drafts, for example), say so rather than reporting a percentage.

6. Improvements ledger

You own Tracking/IMPROVEMENTS.md.

  1. Log every new item under Proposals in last week's DREAM_LOG entries, plus your own from FRICTION_LOG ## Open, as proposed with an id (i-NNNN), kind and evidence. Dedupe.
  2. Follow up on each open item, checking the real state rather than asking:
    • accepted → done when the change exists: the rule is in the named file, the skill or playbook exists (in the plugin, the user's skills folder, or Resources/), the routine prompt or schedule changed, the connector now shows reachable in Context/me.md.
    • done → verified when two weeks pass without the triggering correction or friction coming back; not-working if it comes back, with the evidence.
    • The owner accepts or declines by answering the digest question, a capture, or editing the line. Move declined items to ## Declined with the date.
  3. Ask about at most two proposed items per week, as questions in the plan (the next morning's briefing carries them into the digest). Lead with the one with the strongest evidence.

Changes to any CLAUDE.md are applied by the owner or by /workbench:migrate, never by you. Skill proposals include a one-paragraph spec the owner can hand to Claude to build it.

7. The contradiction queue

Collect every ## Needs reconciling section across the live vault. Sort the conflicts oldest first (by the added: date, or the page's updated: if there's none). Put at most three in the plan under ## Needs reconciling, as numbered questions the next briefing carries into the digest, each with both conflicting claims and the page each one sits on:

Reconcile 1. When does the Acme pilot start? [[Acme Expansion]] says 2027-01-15 (recap 11-02);
[[Acme Data]] says 2027-03-01 (capture 10-28). Which is right?

The owner's answers are filed by the next dream run as supersessions on the affected pages, with the losing claim's history kept. You don't edit the sections yourself. Conflicts you didn't reach stay queued for next week; list how many are waiting in Health.

8. Commitment closure

In Tracking/COMMITMENTS.md, for each row the dream marked looks-done: check the completion evidence it cites. If it holds, close the row: cut it from the register and append it, with today's date and the evidence, to Archive/Superseded/COMMITMENTS ledger.md (create it if missing). Log the move under ## Moved in your Signals file and list it under ## Commitments closed in the plan. Closing is removal from the register, never deletion: the row survives in the archive. This is the only step in any routine that removes register rows. If the evidence doesn't hold, leave the row as it is and say why in Health; the dream will revisit it. Never close an open or looks-stale row.

Finish: the run ledger

Last, append exactly one line to the bottom of Tracking/Signals/RUN_LEDGER.md, and never edit or delete an existing line. Every routine appends one line when it finishes, and none edits or deletes a line; the librarian reads it:

- YYYY-MM-DD HH:MM | weekly | <ok|partial|failed> | wrote: <file names, or "nothing"> | gaps: <none, or a short description>

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.