where-were-we
Skill2 starsChanged 41 days agoMITView on GitHub
Catch the user up mid-session when the conversation has run long and they've lost the thread — a short status read on what this work is for, what has actually landed, where it sits in the flow (did ship-check / mr-review / fix-review run yet?), and what is still unresolved. Fires on two signals. (1) The user has lost track of the session as a whole - /where-were-we, "where were we", "what were we doing", "catch me up", "remind me where we are". (2) The user is at the end of a change and wants a pre-merge gate check on their own process - "did we run ship-check yet", "what haven't we run", "am I clear to merge", "did we skip a step". Do NOT fire on an ordinary in-task question like "what's left on this?" or "what's next" - those want a one-line answer about the current step, not a full status read. Read-only - never resumes the work, fixes anything, writes files, or commits.
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 39
When this fires as a pre-merge gate check rather than a catch-up, this section *is* the answer: lead with it, and keep the other four to a …
- Distinctive: line 56
Nothing open is a real answer: If every loop is closed, say so in one line and give the next action.
- Very specific: line 58
If a saved handover doc exists for this branch in .claude/handover/, mention it in one line and point at /handover-resume — do not read it …
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
The user has lost the thread of a long session. Give it back in under 20 lines. This is a status read, not a story — nobody needs the narrative of how we got here.
Do not summarize the conversation. A recap of everything said is the failure mode; that is what they already couldn't hold in their head. Report state, not history.
Answer five things, in this order, and stop:
Goal — one sentence: what this session is actually trying to land. Take it from the user's original ask, not from whatever the last ten messages drifted into.
Landed — what is genuinely done, one line each. Commits (with short SHAs), pushes, files
written, decisions settled. Cross-check against ground truth: a claim in the transcript that
something shipped is not evidence that it shipped. Run git log --oneline, git status, glab
mr list as needed — if the conversation and the repo disagree, the repo wins and the
disagreement is itself worth a line.
Where we are in the flow — place the work on the chain in the collection README's flow map ("When to reach for which"). That map is the single source of truth — read it, do not restate a chain from memory here. If what you remember disagrees with the README, the README wins. Name the steps that ran and the ones that have not.
Distinguish not yet run from not on the board. If no MR exists, /mr-review is not
pending — it is not applicable yet. Listing inapplicable steps as outstanding invents work.
Steps leave different amounts of evidence — do not treat them alike.
- Provable from ground truth: commits and pushes (
git log,git status), the MR itself (glab mr list --source-branch),mr-reviewandfix-review(both post threads to the MR — check its discussions). - No artifact at all:
ship-check,root-cause,validate-plan,prove-the-test. Nothing on disk proves these ran. If the transcript doesn't show one, say "no evidence it ran" — never assert it didn't, and never assert it did.
When this fires as a pre-merge gate check rather than a catch-up, this section is the answer: lead with it, and keep the other four to a line each.
Still open — the part they actually asked for, and the part that needs discipline. An open loop needs evidence: something I said I would do and did not, a finding nobody accepted or rejected, a question I asked that went unanswered, a check that failed, a caveat raised and never resolved. Uncommitted work counts. "We could also…" does not — an idea nobody raised is not an open loop, and padding this section is how the skill becomes noise.
Next — one line. The single most useful thing to do now.
Rules
- Read-only. Report and stop. Do not resume the work, fix the open items, write a handover doc, or commit. If they want any of that they will say so next.
- Brevity is the feature. They asked because there was too much to hold. Bullets, no preamble, no restating the question back at them.
- Nothing open is a real answer. If every loop is closed, say so in one line and give the next action. Do not manufacture loose ends to fill the section.
- If a saved handover doc exists for this branch in
.claude/handover/, mention it in one line and point at/handover-resume— do not read it in and re-brief from it.
Other versions1 project uses a version of this.
Word for word · 1
- AhmedMerza/claude-skills★ 2100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
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 repository53 other files it gives its agents.
- animateSkill
- api-docs-completeSkill
- best-optionSkill
- browseSkill
- changelog-generateSkill
- checkpointSkill
- claimSkill
- commitSkill
- courtSkill
- design-driftSkill
- designSkill
- expandSkill
- explainSkill
- fix-reviewSkill
- grill-meSkill
- handover-listSkill
- handover-resumeSkill
- handover-saveSkill
- i18n-syncSkill
- issueSkill
- leadSkill
- mr-approveSkill
- mr-assignSkill
- mr-createSkill
- mr-guideSkill
- mr-listSkill
- mr-mergeSkill
- mr-review-requestSkill
- mr-reviewSkill
- mr-statusSkill
- persona-walkthroughSkill
- ponytailSkill
- prove-the-testSkill
- qa-crawlSkill
- qa-sweepSkill
- reflectSkill
- reviewSkill
- root-causeSkill
- scoutSkill
- second-opinionSkill
- ship-checkSkill
- skill-auditSkill
- skill-compareSkill
- spinoffSkill
- story-siteSkill
- tool-compareSkill
- ui-auditSkill
- ui-polishSkill
- usage-weekSkill
- ux-auditSkill
- validate-planSkill
- video-teardownSkill
- wait-whatSkill
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.

