| name | write-tune-up |
| description | Apply the week's learning to the team's written rules, raise everything that would rewire the team as a task card, and write the one-page tune-up report. Use as the closing step of the weekly tune-up. |
| audience | team |
Writing the tune-up
Two steps found what moved and what should change. This step is where a report either
becomes a better team or becomes wallpaper.
1. Apply the left column
For each proposal marked applied — a rule, a rubric line, a verified claim — make the
edit now, in the owner's words where you have them.
- Add the line. Do not restructure the file around it.
- Never delete an existing rule to make room. If a new rule contradicts an old one, that
contradiction is a card for the owner, not a decision for you.
- After each edit, read the new line back in the report so a rule written into the wrong
file is visible immediately.
2. Card the right column
Everything that would rewire the team — a model change, an upstream pull, a connector fix, a
new or retired workflow — becomes one card in tasks/:
---
status: todo
for: <the agent that would own the change, or omit>
---
# <The change, in one line>
**Found:** <what happened, with the count — "failed 3 runs since Aug 18">
**Change:** <the exact edit, file and text>
**If ignored:** <what keeps happening>
One card per change. Never bundle three changes into one card — a bundled card is refused
whole because of the one item the owner disagrees with.
Urgent connector findings jump the queue: a connector that is used but unreachable gets
its card written first, because the next scheduled run will fail at 6am whether or not
anyone read this report.
3. Write the report
The workflow's output path. One page. The owner reads this on a phone on a Sunday.
# Weekly tune-up — <date>
## Report card
**Made:** <n> rules applied, <n> cards raised, <n> things checked.
**Quality:** n/a — this is the tune-up itself
**Confidence:** high | medium | low — <one sentence, lower it when a lane could not be checked>
**Sources:** runs/ (<n> logs), quality/verdicts/ (<n>), tasks/ (<n> open), <lanes checked>
When you count `tasks/`, **skip `README.md`** — it explains the folder and is not somebody's
card. `work-the-tasks` skips it and so does the dashboard; counting it here would report one
more open task than either of them shows, every single week.
**Needs you:** <the single most important card, or "nothing">
## Did last week's proposals get done?
—
—
—
—
4. "Nothing happened twice" is a real answer
A quiet week gets a short report that says so. Do not manufacture findings to fill a
template — an invented pattern becomes a rule, and a wrong rule makes every future draft
worse while looking like the system is working.
The Not checked section is never omitted either. A lane that silently disappears from
the report is how a team stops noticing it went blind.
5. Finish like every run
Follow .claude/skills/run-log/SKILL.md. The report, the applied edits, and any cards go in
the same commit. This run carries no quality block — the tune-up is not graded, it is the
thing that improves what gets graded.
If you were told not to commit, do not commit — leave the report, the edits and the cards
uncommitted together. What matters is that they move as one: applied edits with no report
explaining them are changes nobody can trace.
git add inbox/<date>/tune-up.md shared/writing-rules.md tasks/ agents/orchestrator/output/last-checked.md runs/<YYYY-MM>/<run_id>.json
git commit -m "tune-up: 2 rules applied, 1 card raised for <date>"
