Przejdź do treści
Biblioteka skilli
Zarządzanie i produkt · Analiza

summary

/summary

Raportuje stan całej partii pracy: przegląda na żywo PR-y, MR-y i zgłoszenia, grupuje je po ticketach i podaje linki.

Repozytorium
corgi ↗
Licencja
MIT
Ostatnia zmiana repozytorium
26.10.2026

Ocena i bezpieczeństwo

100/ 100Klasa: dobry
Format25 / 25
Frontmatter z nazwą i opisem, poprawna nazwa, opis do 1024 znaków, nazwa zgodna z folderem.
Odkrywalność25 / 25
Opis mówi, kiedy użyć skilla, ma sensowną długość i jest w 3. osobie.
Treść35 / 35
Co najmniej 150 słów i do 500 linii, nagłówki, listy kroków, przykłady albo kod.
Higiena15 / 15
Bez zaślepek typu TODO, bez ścieżek Windows, z licencją.
Ryzykobrak sygnałów w skanie
Pliki w paczce
1
Skrypty
0
Komentarze HTML
0
Adresy URL w SKILL.md
2
Przestarzałość

Brak sygnałów. Repozytorium zmienione 26.10.2026, bez odwołań do wycofanych modeli i API.

Licencja

MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.

Co robi i kiedy

Raportuje stan całej partii pracy: przegląda na żywo PR-y, MR-y i zgłoszenia, grupuje je po ticketach i podaje linki.

Wyzwalacze z opisu

give me a summary\what, \
Oryginalny opis (EN)

Use when someone asks where a batch of work stands: \"give me a summary\", \"what's left\", \"status of all the MRs\", \"what did we ship today\", \"recap\", \"which PRs are still open\". Reads the live state of every PR/MR and ticket the session touched and reports it grouped by ticket, with clickable links. NOT a code review (review), NOT a standup post for someone else (standup).

Jak działa

  1. Overview
  2. Guardrails
  3. Phase 1 - Gather, live
  4. Phase 2 - Group by ticket, not by repo
  5. Phase 3 - Say what mattered
  6. Phase 4 - Check it before sending
  7. Red flags - stop
  8. See also
Sekcje z nagłówków SKILL.md, w oryginale · 123 linii, 1 011 słów
Podgląd SKILL.md
---
name: summary
description: "Use when someone asks where a batch of work stands: \"give me a summary\", \"what's left\", \"status of all the MRs\", \"what did we ship today\", \"recap\", \"which PRs are still open\". Reads the live state of every PR/MR and ticket the session touched and reports it grouped by ticket, with clickable links. NOT a code review (review), NOT a standup post for someone else (standup)."
---

# Where the batch actually stands

## Overview

After a long session the state lives in three places that disagree: what the session
remembers doing, what the forge says now, and what the tracker says now. Auto-merge
fires, someone else merges a branch, a pipeline goes red an hour after it went green.
A summary written from memory is wrong by the time it is read.

So: **re-read the live state, then report it grouped the way the work was asked for** -
one block per ticket, not one flat list of links.

Neighbours:

| Skill | For |
| --- | --- |
| **this one** | the user asking where their own batch stands, now |
| `standup` | a post for the team channel, written for people who were not here |
| `review` | judging the code in a PR, not reporting its state |
| `tracker` | moving tickets, not describing them |

## Guardrails

- **Never report state from memory.** Every row is re-fetched at summary time. "Merged"
  because the session armed auto-merge two hours ago is a guess, and the one row that is
  wrong is the one that matters.
- **Links must be clickable.** Markdown links in a table. A fenced code block renders
  monospace and kills the links - if a user asks for aligned columns, the table is the
  answer that keeps both.
- **Say what is not done.** The open rows and the blocked rows are the reason anyone
  asked. Leading with the merged list buries them.
- **Do not claim a deploy you did not verify.** Merged is not deployed; a manual
  production job is not a deploy. Check, or say "merged, deploy is a manual job".
- **Attribute review findings honestly.** A defect a reviewer caught is theirs, not a
  thing that "was improved" in the passive voice.

## Phase 1 - Gather, live

Collect in one pass, in parallel where the tool allows:

| Source | What to read |
| --- | --- |
| forge | per PR/MR: `state`, head `pipeline.status`, `reviewers`, `merge_when_pipeline_succeeds`, unresolved discussion count |
| tracker | per ticket: current status, assignee |
| the session | which PR/MR belongs to which ticket, and what was deliberately left out |

Sources:
- **GitLab** - `glab mr view <n> -R <repo> -F json`; threads via `…/discussions`.
- **GitHub** - `gh pr view <n> --json state,statusCheckRollup,reviews,reviewRequests`.
- **Tracker** - the `tracker` skill's read path.

A pipeline reading `success` while a job inside it failed under `allow_failure` is
still worth a word; so is an MR whose auto-merge was dropped by a later push.

## Phase 2 - Group by ticket, not by repo

One block per ticket, headed with its key and a short plain title, then a table of the
repos that carry it. A ticket spanning four repos is one block with four rows - that is
the shape the work actually has, and a flat list hides which repo is holding it up.

```markdown
## [ABC-123] Short plain title - DONE ✓

| Repo | MR | Reviewer | State |
|---|---|---|---|
| api | [!42](https://…/merge_requests/42) | Sam | MERGED ✓ |
| web | [!37](https://…/merge_requests/37) | Alex | OPEN - auto-merge armed *(waiting on CI)* |
```

Rules for the table:
- **MR/PR is a link**, written `[!42](url)` / `[#42](url)`.
- **State carries the nuance** - `MERGED ✓`, `OPEN - auto-merge armed`, `OPEN - 2
  unresolved threads`, `OPEN - pipeline red`, `CLOSED *(superseded by !44)*`. The
  parenthetical goes in italics in the cell; never in a trailing note the eye skips.
- **Reviewer is who is actually on it**, or `-`. Not who you hoped would take it.
- Mark the ticket's own outcome in the header (`- DONE ✓`) only when the tracker says so.

Then a one-line tracker roll-up under the blocks:

```
Tickets: ABC-123 → Done · ABC-124 → In Review
```

## Phase 3 - Say what mattered

After the tables, at most a few short sections. Include a section only when it changes
what the reader does next:

- **What still needs them** - open decisions, missing credentials, a manual job someone
  has to play, a question waiting on a person. This goes near the top if anything is in it.
- **What a review caught** - defects found in the session's own work, in one line each.
  This is the highest-signal part of most summaries and the easiest to omit out of
  vanity; include it, credited to whoever found it.
- **Anything deliberately not done** - scope consciously left out, with the reason. A
  reader who does not see it assumes it was forgotten.

Skip praise, skip restating each diff, skip a "next steps" list that repeats the open
rows. If everything merged and nothing needs them, the summary is the tables and one line.

## Phase 4 - Check it before sending

- Every link opens the thing it names.
- Every state matches what was fetched this minute.
- The open and blocked rows are visible without scrolling past the merged ones.
- Nothing claims a deploy, a release or an approval that was not observed.

## Red flags - stop

- A row whose state came from what the session did, not from a fetch → re-read it.
- The whole summary in a fenced code block → links are dead; use a table.
- Merged rows first and the blocker at the bottom → reorder.
- "Improved", "cleaned up", "fixed some issues" → name the defect or drop the line.
- A ticket marked Done in the summary but not in the tracker → one of them is wrong;
  find out which before reporting either.

## See also

- `tracker` - reading and moving ticket state.
- `standup` - the same facts written for a team channel.
- `stories` - produces the batch this reports on; its Phase 6 report is this format.
- `ship` - release state, when the question is "is it out" rather than "is it merged".

Wersje

RepozytoriumWynikLicencja
Ta strona · andriiklymiuk/corgi100MIT
optimiziramsi/skills100MIT
JunLin615/os-optics-kit100MIT
yarikleto/claude-teacher-plugin100MIT
brsbl/ottonomous100MIT
brsbl/ottonomous100MIT
alaliqing/claude-paper100MIT
minhuw/claude-writer100MIT

Migawka bazy z 13.11.2026 · wersja paczki 903930b905b1 (skrót SHA-256 całej paczki)

Pobrania w ostatnich 30 dniach: 0 · ranking

Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.

Podobne skille

  • 100standup-summary
    Przygotowuje podsumowanie standupu z ostatniej aktywności: commitów, Jiry, MR, kalendarza, Slacka i zadań z pamięci.
  • 100standup
    Tworzy aktualizację standupu lub dziennik pracy z aktywności, gdy użytkownik pyta, nad czym pracował dzisiaj.
  • 90standup
    Zbiera ostatnią pracę nad zgłoszeniami z artefaktów i historii Git i przygotowuje wczoraj, dziś i blokery na standup.
  • 90standup
    Przygotowuje krótkie podsumowanie na standup: co zrobiono wczoraj, co będzie dziś i jakie są blokery.
  • 100standup
    Przygotowuje dzienne podsumowanie standupu: co ukończono ostatnio, co jest w toku i co jest zablokowane.
  • 100standup-summary
    Tworzy podsumowanie standupu lub tygodnia z commitów, gałęzi, PR i zgłoszeń, zapisane zwykłymi zdaniami, nie żargonem commitów.

Uwagi i zgłoszenia

Wiadomość trafi do Bartosza Sokolińskiego. Adres e-mail jest potrzebny tylko do odpowiedzi. Szczegóły w polityce prywatności.