resume
Skill2 starsChanged 4 months agoMITView on GitHub
Resume a project in five minutes - read BRIEF/PLAN/JOURNAL plus git state, deliver a short where-we-are brief, divergence flags (uncommitted work, unjournaled commits, acceptance drift), and the next three unblocked tasks. Use at session start or when the user asks where were we, what's next, to catch up on a project, or to get oriented in an unfamiliar repo.
58 other projects use a version of this.
Adopted by
Most recently taken up by
- FabianoArthur/claude-code-kitchentheir own version, first seen Nov 8, 2026
- leinardi/adversarial-review-looptheir own version, first seen Nov 8, 2026
- Dariusplayer09/adaptive-skillstheir own version, first seen Nov 8, 2026
- ArcticFox2029/chamnantheir own version, first seen Nov 5, 2026
- inkatze/planwrighttheir own version, first seen Nov 4, 2026
- vzhukovs/podman-desktop-kittheir own version, first seen Nov 1, 2026
Used with
Projects with it also use pause, status and 10 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.
kennethleungtySkillpause
Pause — stop the music (alias for /stop)
Also in 7 projects that use this
AnthropicSkillstatus
Case status summary by audience — client-facing (plain language), internal (for the professor), or court-ready (formal caption format per local rules). Same facts, different framing and depth. Use when a student needs to update the client, brief the professor, or prepare a court status report.
Also in 15 projects that use this
MicrosoftSkillreport
Turn an exploration (threads, findings, charts) into a single Markdown report — note, blog post, executive summary, KPI dashboard, slide brief, or multi-section analytical report, with embedded charts.
Also in 7 projects that use this
MicrosoftSkillconfig
Show or change TokenEconomizer's settings: turn micro-skills, tool pruning or the loop guard on or off, or tune them, for this repository or (with --global) every repository. Use when the user asks to configure, enable or disable TokenEconomizer or one of its mechanisms.
Also in 3 projects that use this
sahinSkillexecute
Also in 4 projects that use this
u9401066Skillbug-fix
Structured bug fixing workflow with root cause analysis and regression prevention. Triggers: BF, 修 bug, fix bug, debug, 除錯, 錯誤, error, 問題, issue, 修復, fix, 故障, fault, 異常, exception, 失敗, failure, 不能用, broken, 壞掉.
Also in 3 projects that use this
- +2 more from microsoft
Other versions58 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 58
- H-mmer/pentest-agents★ 815same name
- bitwize-music-studio/claude-ai-music-skills★ 589same name
- EmeaAppGbb/spec2cloud★ 100same name
- thiagofinch/mega-brain★ 95same name
- saksham10arora-dotcom/everything-claude-codex★ 74same name
- thesun4sky/jobstack★ 50same name
- junjiezhou1122/ideascientist-claude-code★ 35same name
- nimadorostkar/Claude-Skills-collection★ 33same name
- wrg32786/aigent-os★ 25same name
- Hassaan146/claude-skills★ 18same name
48 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 repository27 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- auditSkill
- challengeSkill
- contract-checkSkill
- depsSkill
- design-auditSkill
- designSkill
- doSkill
- eval-runSkill
- eval-specSkill
- healthSkill
- investigateSkill
- journalSkill
- learnSkill
- migrate-checkSkill
- plan-reviewSkill
- planSkill
- qaSkill
- refactorSkill
- retroSkill
- secureSkill
- shipSkill
- ticketSkill
- triageSkill
- verifySkill
- whySkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 4 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 51 edited versions we read keep these.
4
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Run before the skill's steps — per invocation, not per session (4.36).
- Very specific: line 37
Never delete a decision. Supersede it: ~~old~~ → Verdict (YYYY-MM-DD): new call — reason.
- Distinctive: line 38
Never fix, tune, or delete a test or eval case to raise a score.
- Distinctive: line 78
No journal commit in history → say so and treat the log since the last dated JOURNAL.md entry (or, failing that, the whole log) as …
- Very specific: line 122
Before listing a candidate, look for a claim signal and print it beside the ID: a remote branch under the project's branch-prefix carrying …
- Distinctive: line 156
Read-only here too: no issue edits, no labels, no comments.
“Common” is measured: kept by at least 6 in 10 of the 51 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
/resume — resume in five minutes
You are rebuilding context, not doing work. Read the three documents and the git state, say where the project actually is, and stop.
Adjacent skills: /journal (writes the session record at session end;
/resume reads it at session start) · /do (starts a task; /resume only names
the candidates) · /triage (full plan/backlog hygiene sweep; /resume flags
only what it trips over while reading).
<!-- acstack:runtime -->Run before the skill's steps — per invocation, not per session (4.36); failures degrade to markdown:
link="$(readlink "$HOME/.claude/skills/health" 2>/dev/null || true)" # empty = not symlinked
pack="$(dirname "$(dirname "$link")")" # NEVER trust this unless $link was non-empty
if [ "${link#/}" != "$link" ] && [ -x "$pack/bin/acstack-config" ] && ! "$pack/bin/acstack-config" runtime | grep -q '=off'; then
"$pack/bin/acstack-config" || true # resolved keys, with sources
"$pack/bin/acstack-update-check" || true # ≤1 fetch/day; silent ONLY if already checked today
"$pack/bin/acstack-recall" || true # LEARNINGS.md + bug-class names, capped 3KB
else
echo "runtime off — proceeding without recall/update-check"
fi<!-- /acstack:runtime --><!-- acstack:principles -->Operating principles
- Be direct. Push back in writing when the plan or the user is wrong. No sycophancy.
- Never delete a decision. Supersede it:
~~old~~ → **Verdict (YYYY-MM-DD):** new call — reason. - Never fix, tune, or delete a test or eval case to raise a score. Log the miss honestly and leave the case unchanged.
- Name exact things: regex patterns, function signatures, model names, before → after numbers. Never "fixed bugs".
- Attribution: follow the project's
attributionsetting (defaultnone) — no AI-tool mentions in generated docs, no attribution trailers in commits or PRs. Commit with explicit-m/-Fmessages only. - Config: read
.claude/acstack.mdat the project root (fall back to~/.claude/acstack.md) before acting.## Settingskeys override pack defaults; a## <skill-name>section overrides both. Unknown keys and sections are ignored. - Docs: BRIEF.md (frozen seed) / PLAN.md (living plan) / JOURNAL.md (rolling journal). If the repo uses legacy names (PLANNING_PROMPT.md / PLANNING.md / STATUS.md), use those instead — never create both.
- Recall: if
LEARNINGS.mdexists at the project root, read it before starting. - Conduct: follow the
acstack-conductblock in this repo's AGENTS.md — the word is the mode; the user sets the pace. - Hackathon lane: if the project's AGENTS.md carries the
acstack:hackathon-laneblock, only/dochanges the repository during the event. Any other skill that would write a tracked file, commit or push says what it would have done and stops; a change that is not a task goes through the lane's operator route.
<!-- /acstack:principles -->One document set. Resolve exactly ONE BRIEF/PLAN/JOURNAL set and name
its path in the report's scope line. If more than one candidate set exists
— a monorepo, nested products, an apps/* tree each with its own docs —
list the candidates and STOP. Never pick one silently: a confident answer
about the wrong product is worse than no answer (conduct rule 8).
What to read
- Config (per the principles block) — note
trackingandjournal-commit-format. - BRIEF.md, PLAN.md, JOURNAL.md (legacy names per the principles block).
A missing document is reported as a fact and pointed at
/plan— /resume never creates or scaffolds anything. All three missing → this is an unfamiliar repo, not a broken one: switch toreferences/mode-cold.mdrather than reporting an empty brief. Retrieve, don't ingest. Once JOURNAL.md is past roughly 500 lines, stop reading it whole: read its blockquote and TL;DR, then its###entry headings, then fetch the FULL text of only the newest entry (plus any heading that matches what the user asked about). A five-minute catch-up that spends its budget loading six months of history has already failed at the thing it is for, and a whole-file read of a long journal crowds out the plan and the git state — the two inputs the brief actually needs. Say which entries were read in full. git status, andgit logsince the most recent commit whose subject matches the journal commit format (defaultJournal <date>: <summary>). Match the format as a PREFIX, not in full. A second entry on one day is legitimatelyJournal 2026-07-29 (3rd): …, and a full-string match misses it — which made this skill count six journal commits as unjournaled on its own repo. Anything starting withJournal <date>is a journal commit. No journal commit in history → say so and treat the log since the last dated JOURNAL.md entry (or, failing that, the whole log) as unjournaled.
The brief — 10 lines or fewer
- What this project is: one line, from the BRIEF.
- Current phase, its
**Exit criterion:**, and its checkbox state. - What the last session actually did — from the newest JOURNAL entry, with
its numbers (
88 passed, 40 skipped), not a paraphrase. - The state of the tree — clean or not, ahead/behind — as the canonical scope
line, one source line, because everything below is about this worktree
(5.17.5):
Scope: branch
<branch>@<sha>vs<default>@<sha>— a verdict about this branch's tree, not the project's; the merged tree is the integrator's to re-check.
Divergence flags
Report only what the reads above surface — this is a spot check, not an audit (a full sweep is /triage's job):
- Uncommitted changes, by file.
- Commits made since the last journal entry — named as candidates for
/journal, not journaled on the spot. - Any checked box whose
**Acceptance:**command is cheap to run and visibly fails now. Run only cheap checks (a grep, anls, a file-exists test — not test suites); say which acceptance commands were NOT run.
Next 3 unblocked subtasks
From PLAN.md in plan order: ID, task text, and its acceptance line.
Unblocked = every prerequisite box checked and not blocked by an open
decision in ## Open items. Fewer than three exist → list what's there and
say why the rest are blocked. Never pad the list to three. A blocked
item does not become ready work by being annotated: listing one under a
heading that says "unblocked" is a false ready-signal, and "three" is a cap,
not a quota. This applies in both modes — tickets mode below relies on
this paragraph rather than restating it (4.75).
A task carrying no **Acceptance:** line
is still listed, flagged as no acceptance recorded — the sibling of
/do's stop on the same gap. Do not invent one to fill the column: a task
whose done-condition is unwritten is a real finding about the plan, and
/plan-review or /triage is where it gets fixed.
Two sessions from one base get the same three (5.17.5). The list is a
function of this worktree's PLAN, so it is a proposal, never an assignment.
Before listing a candidate, look for a claim signal and print it beside the
ID: a remote branch under the project's branch-prefix carrying the task ID,
and an open PR mentioning it —
git for-each-ref --format='%(refname:short)' "refs/remotes/origin/<branch-prefix><id>*"
and gh pr list --state open --search "<id>". A hit keeps the candidate
listed, flagged may already be taken: <branch or PR>; no hit prints
no claim seen as of the last fetch — /resume never fetches, and a session
that has not pushed is invisible, so silence is not proof. No remote or no
gh → claim check skipped — <why>, never a silent omission.
Mode: cold — no doc triad
Full procedure: references/mode-cold.md — read it when BRIEF/PLAN/JOURNAL
are all absent, and not otherwise. It replaces the document reads with
README, manifest, layout, shaped git history and CI, and it forbids the
two things a stranger's repo punishes: scaffolding, and a "next 3 tasks"
list invented from nothing.
Tickets mode (tracking: tickets)
Preconditions first (gh present, authenticated, GitHub remote); a failure is named exactly and the document-side reads above still happen. Then the checkbox scan becomes a tracker query:
gh issue listopen issues in the current milestone; milestone burn stated as open/closed counts against its exit criterion.- Next 3 = top unblocked open issues (no
blockedlabel) in the current milestone, each with its acceptance section's first line and its assignee — an assigned issue is taken (read it, never set it; 5.17.5). The fewer-than-three rule in "Next 3 unblocked subtasks" above governs here too — three is a cap, not a quota, and ablockedissue is never listed to reach it. This section omitted that clause until 2026-08-14; with one unblocked issue in the milestone, a live run padded to three by including theblockedone under an "unblocked" heading (4.75). - Read-only here too: no issue edits, no labels, no comments.
Hard rules
- /resume writes no files and changes no state — in either mode.
- End with a status statement. Naming the next tasks is information, not an offer accepted: work starts only when the user picks something (CONDUCT rules 2 and 9).
- What was not checked is stated plainly — a five-minute brief that implies it verified everything is lying about its own depth.
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.

