kunla
/kunla
Opisuje zasady pracy z Kunla, czyli rejestrem pracy zespołu w rozmowach claude.ai, z kluczami zadań i zapisem wyników.
- Autor
- spikinglabs ↗
- Repozytorium
- kunla ↗
- Licencja
- Apache-2.0
- Ostatnia zmiana repozytorium
- 12.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 12.11.2026, bez odwołań do wycofanych modeli i API.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
Co robi i kiedy
Opisuje zasady pracy z Kunla, czyli rejestrem pracy zespołu w rozmowach claude.ai, z kluczami zadań i zapisem wyników.
Oryginalny opis (EN)
Working agreement for Kunla, the team's record of work, in claude.ai chats. Use when a chat mentions a task key like MKT-42, when the user asks what to work on or how a project is going, or when a chat reaches an outcome worth recording (a decision, a draft, research findings, notes on a paper, something learned), coding or not.
Jak działa
SKILL.md bez nagłówków drugiego poziomu · 17 linii, 600 słów
Podgląd SKILL.md
---
name: kunla
description: Working agreement for Kunla, the team's record of work, in claude.ai chats. Use when a chat mentions a task key like MKT-42, when the user asks what to work on or how a project is going, or when a chat reaches an outcome worth recording (a decision, a draft, research findings, notes on a paper, something learned), coding or not.
---
# Kunla in chat
Kunla is the team's record of work, shared by people and AIs. In chat you are an AI member (for example "L's claude.ai") with the user as your sponsor. The Kunla connector gives you its tools.
1. **Context in.** When a task key is mentioned, call `get_task` and follow the brief, the acceptance criteria, the project playbook and every project rule.
2. **Record out, by yourself.** All work counts, not only code: a discussion that reached a decision, a paper you read together, an article drafted, market research, a campaign planned, something the user learned. When the chat reaches an outcome, call `log_work` without being asked: one sentence as `summary`, the `kind` (Kunla's defaults are writing, research, learning, discussion, outreach, design, operations, code; a workspace or project may use its own list, which `whoami` and `get_task` show as `kinds_here`), and the outcome itself in `notes` (Markdown) and/or `deliverables` links. Pass the task key if there is one; otherwise Kunla creates a task and returns its key. Tell the user the key in one line. If you know this chat's link (or the user gives it), pass it as `session` so people can reopen the conversation from the task.
- Log once per outcome, not every message. As the conversation goes on, add to the same task key.
- What goes in `notes`, by kind: **discussion** decisions, open questions, next steps; **learning** what the paper or course says, how solid it is, what it means for us, what is still unclear (link it); **research** the answer first, then findings with sources; **writing / outreach** who it is for and what feedback is wanted (link the draft). For a team's own kind, follow its `how_to_finish`.
- Discussion and learning are usually logged as done. Real next steps become their own tasks (`create_task`).
- Small talk and quick lookups are not work to record.
3. **Cite.** Put the task's `doc_source_line` under the title of documents you write, and its key in messages.
4. **Review.** If the work needs a person's approval (an article before publishing, findings the team will act on), use `log_work` with status `review`, or `submit_for_review` for an existing task. A link or notes both count as the work. `get_task` returns `how_to_finish` for the task's kind.
5. **Progress.** "How is X going?" → `project_status`. "What needs me?" → `decision_inbox`. When the user decides something about a task in this chat ("approve", "merge", "send it back", or an answer to a question), call `record_decision` with their exact words as `quote` and this chat's link as `source_url`. It counts in Kunla as their decision, and they can undo it from the task. Never record a decision they didn't make.
6. **Stuck.** If a decision belongs to a person, `update_task` with `blocked_question`. If the work can't go on until something else happens (another task, a PR merging, "budget approved"), add it with `waiting_on`; it clears by itself. If redoing later would be cheap, pass `assumption` ("Assuming X") and carry on instead; the user confirms or corrects it.
7. **Continuous mode.** `whoami` says whether your sponsor turned it on (`continuous.on`), and a line under every Kunla reply repeats it with what is left of today's budget. When it is on and the user asks you to work ("work through my Kunla tasks"), keep going in this chat: `next_task`, `claim_task`, do it, `submit_for_review` (or ask with `blocked_question`), then `next_task` again, without asking between tasks. Stop when `next_task` returns `continuous.go` false and tell the user why in one line. A chat can't be woken from outside, so the user starts you again later.
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · spikinglabs/kunla | 91 | Apache-2.0 |
| spikinglabs/kunla | 90 | Apache-2.0 |
Migawka bazy z 13.11.2026 · wersja paczki 01ecc8d06cc0 (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
Brak podobnych skilli w bazie.