daily-review
Skill4 starsChanged 2 months agoApache-2.0View on GitHub
Read the daily loop's review page, talk the proposed specs through with the operator, and record their decisions so tonight's build window can act on them. Publishes the review page as an Artifact, answers questions about any proposal, applies requested edits to openspec/changes/, and writes spec-queue/APPROVALS.md. Use when the operator says "today's review", "daily review", "review the specs", "what did the loop propose", "approve today's specs", or opens a session after the day window has run. This is the human step of the cycle in openspec/explorations/2026-09-01-a-daily-research-spec-and-build-loop.md; the two scheduled windows either side of it are unattended and cannot ask anything.
18 other projects use a version of this.
Adopted by
Most recently taken up by
- mrosano1987/git-daily-reviewtheir own version, first seen Nov 4, 2026
- AN0099/toothpaste-kittheir own version, first seen Oct 24, 2026
- RumbleJack56/skillstheir own version, first seen Oct 20, 2026
- razbakov/skillstheir own version, first seen Oct 17, 2026
- bradautomates/second-braintheir own version, first seen Oct 9, 2026
- zephyrwang6/myskilltheir own version, first seen Oct 8, 2026
Used with
Projects with it also use weekly-review.
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.
QuixiAISkillweekly-review
Review the past week across goals, tasks, calendar, journal, and commitments, then propose the next week
Also in 3 projects that use this
Other versions18 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 18
- heyitsnoah/claudesidian★ 2.6ksame name
- qusong0627/QuantMind★ 1.9ksame name
- davekilleen/Dex★ 509same name
- zephyrwang6/myskill★ 339same name
- bradautomates/second-brain★ 217same name
- SpaceZephyr/allSkills★ 14same name
- timominkus/obsidian-para-gtd-ai★ 9same name
- razbakov/skills★ 5same name
- finger88/my-claude-skills★ 3same name
- RumbleJack56/skills★ 2same name
8 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- autonomous-prepSkill
- autonomous-sessionSkill
- backlogSkill
- check-buildSkill
- copilot-test-setupSkill
- e2e-loopSkill
- handoffSkill
- harness-auditSkill
- harness-refreshSkill
- night-briefingSkill
- openspec-apply-changeSkill
- openspec-archive-changeSkill
- openspec-exploreSkill
- openspec-proposeSkill
- openspec-sync-specsSkill
- resumeSkill
- review-iterationSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 6 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 11 edited versions we read keep these.
6
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
- Distinctive: line 48
Keep the same artifact URL across days if the operator wants one running page.
- Distinctive: line 65
When the operator wants a change different:
- Distinctive: line 96
Commit it. Then tell the operator, in one short list, exactly what tonight's window will build — including the backlog items that come …
- Distinctive: line 104
Check git branch --merged master for the cycle branch.
- Distinctive: line 113
Never mark a task complete on the strength of a plan existing.
- Distinctive: line 114
Do not touch port 8000. 8010 is the trial Hub.
“Common” is measured: kept by at least 6 in 10 of the 11 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.
Files
The daily loop's one human step. Everything either side of it runs unattended and cannot ask a question, so this is where judgement enters.
Contract for every file named here: spec-queue/README.md. Playbooks for the windows either side:
.claude/loops/.
Step 1 — Find out what happened, before opening anything
powershell -NoProfile -Command "Get-Date -Format 'yyyy-MM-dd'" # Git Bash date is skewed here
ls -t spec-queue/review/*.html | head -3
git branch --show-current
git log --oneline -15
git status --shortThen read, in this order:
- The newest review page in
spec-queue/review/. If it is not today's, say so in your first sentence — the operator is about to make decisions and needs to know they are about a day they may not remember. - The bottom of the newest
.claude/autonomous/*-day-log.md, for what the window actually did versus what the page claims. spec-queue/APPROVALS.md— is there already a section for today? Someone may have started.
If there is no review page at all, do not improvise one. Report what you found — whether the day
window ran, what its log says, whether AgentWeaveResearch produced a file in
~/.claude/routines/agentweave-research/out/ — and stop. A review page invented here has not been
through the rounds and must not reach APPROVALS.md.
Step 2 — Publish it
Publish the review page as an Artifact and give the operator the link. This is the reason the human
step exists in an interactive session at all: headless claude -p has no Artifact tool
(verified 2026-08-28; no CLI flag provisions one), so the window that wrote the page could not
publish it.
The page is already self-contained and theme-aware. Strip the outer
<!DOCTYPE>/<html>/<head>/<body> wrapper before publishing — the Artifact runtime supplies its own
skeleton and would nest them — but leave the file on disk intact, since it is also meant to open in
a browser.
Keep the same artifact URL across days if the operator wants one running page; use a fresh one per day if they would rather have the history. Ask once, then stop asking.
Step 3 — Talk it through
This is the part that cannot be automated, so do not rush it into a checklist.
For each proposed change, the operator needs enough to decide and no more: what breaks today, what the change does about it, what it touches, and its tier, its acceptance drive, and what the review round changed. That last one is the tell. A change where two independent re-derivations found nothing is either genuinely clean or was not really re-derived, and the operator is entitled to know which the page claims.
Answer questions by reading the actual proposal and the actual code — openspec/changes/<name>/ and
the files it names. Do not answer from the review page alone; it is a summary written by a process
that was optimising for brevity.
When the operator wants a change different:
- Small correction — apply it to
openspec/changes/<name>/now, re-runopenspec validate --strict <name>, and mark the rowAPPROVED. Say in the note what you changed. - The argument is wrong — mark it
REVISINGwith what needs re-deriving. Tomorrow's day window takes another round. Do not rewrite an argument here and then approve your own rewrite; that is the round discipline collapsed to one round, performed by the person who is meant to be checking it. - Not now —
REJECTED, with the reason. The reason matters: without one it will be re-proposed.
Step 4 — Write the decisions down
Append today's section to spec-queue/APPROVALS.md, newest day first, using the format in
spec-queue/README.md. The status token is the authority; there is no checkbox.
## <today>
- APPROVED <change-name> note
- REVISING <change-name> what needs re-deriving
- REJECTED <change-name> whyOffer the review page's Tier-0 fix list and write the operator's picks as one
APPROVED-FIXES: F<n>, F<n> line. Add ORDER: only if the operator asked for a different order tonight. Add NOTHING TONIGHT if they
want the window to stand down — it stops before spending a model invocation.
Leave undecided changes out. Absence is not rejection: a row-less change stays in
openspec/changes/ and comes back in tomorrow's page. Inventing a decision the operator did not make
is the one failure this step cannot recover from, because the night window treats this file as
authoritative and nobody is awake to contradict it.
Commit it. Then tell the operator, in one short list, exactly what tonight's window will build — including the backlog items that come first by default, so the order is never a surprise in the morning.
Step 5 — The merge question
The cycle branch is merged by the operator, awake. Never automatically, and never by a window.
Check git branch --merged master for the cycle branch. If it is unmerged and spans more than a day,
raise it — an accumulating diff that nobody merges is worse than no loop, because it eventually grows
past reading. Offer the merge; do not perform it unasked.
Limits
- Push, do not open PRs.
- Stage explicit paths, never
git add -A. openspec validate --strict <name>must pass before any row is markedAPPROVED.- Never mark a task complete on the strength of a plan existing.
- Do not touch port 8000. 8010 is the trial Hub; the windows' drive Hub uses a port chosen that night, which the window's log records.
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
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.

