Pull

Never write to the tracker. Run the steps in order; stop and ask when a step says so.

0. Before anything

1. Desktop. If no mcp__ccd_session__start_session tool exists (loaded or deferred), send error.desktop-only (in the user's language) and stop. Create nothing.
2. Profile. Read ~/.config/rabbit/profile.md (its Workflow: line is not used here). If it is missing or unreadable, or a required value is empty (Basics: Workspace, Repositories folder, Chat language, Document language; Tracker: Site, Tools, Sprint field), send setup.needed naming what is missing, run the rabbit:setup skill with the Skill tool, and when it finishes continue here. Site or Tools still empty after setup: stop (setup has said what to connect). Sprint field still empty (no sprints on this site): send error.no-sprint and stop.
3. Language. LANG = Chat language, or for auto the language of the user's last message (English if none). DOC_LANG = Document language, or LANG for auto. Every message stays in LANG, even after text in another language. Send fixed messages from references/messages.md by id, translated; keep Markdown, {{...}} values, paths, keys and commands as they are.

Constants: FIELD = Tracker Sprint field; tracker calls use the profile's Tools (with cloudId = Site when the tool takes one); WORKSPACE = Basics Workspace with ~ expanded; sidebar section Sprints.

1. Read the sprint

Search with jql: "assignee = currentUser() AND sprint in openSprints() ORDER BY key", fields: ["summary", "status", "issuetype", "priority", "labels", "parent", "description", FIELD], maxResults: 100. Never ask for an "evidence" or expanded view. Follow the tool's paging (for example nextPageToken until isLast, or startAt) until every issue is read.

From each issue's FIELD keep the active entries: state active in any case, or a string entry such as …Sprint@x[id=12,…,state=ACTIVE,name=…] (Jira Server/Data Center), read the same way. No issue: send error.no-sprint and stop. If they span more than one sprint id, send ask.which-sprint with the list, end the turn, and keep only the chosen sprint's issues. SPRINT_ID = that sprint's id; LABEL = SP_<SPRINT_ID>; SPRINT_DIR = "<WORKSPACE>/<LABEL>" (always quote paths).

Unfinished keys: the same JQL with AND statusCategory != Done before ORDER BY (fields: ["summary"]), kept only when the key is among the chosen sprint's issues. Do not infer status categories from status names.

2. Become the sprint's main session

1. mcp__ccd_sidebar__list_sections: take the first section named Sprints; none: create it with mcp__ccd_sidebar__create_section.
2. mcp__ccd_session_mgmt__list_sessions with limit: 500 (this session is never in the result).
3. A session titled exactly LABEL means another session is this sprint's main session: send error.other-main linking it as [<title>](#<sessionId>) and stop. Create nothing.
4. Otherwise move this session into the section (mcp__ccd_sidebar__move_sessions, session_ids: ["self"], section_id). If refused because this session is nested, run mcp__ccd_session_mgmt__detach_session with "self" and move again. Title it LABEL (mcp__ccd_session_mgmt__set_session_title, "self").
5. mkdir -p SPRINT_DIR and mcp__ccd_directory__change_directory to it (it applies after this turn; until then use absolute paths).

3. Decide which tasks need a session

mcp__ccd_session_mgmt__get_session with "self" gives this session's id; mcp__ccd_session_mgmt__list_sessions with linked: true, include_archived: true, limit: 500 lists the sessions it started. A task already has a session when one of them has startedBy equal to that id and a title matching (^|[^A-Z0-9-])<KEY>([^0-9]|$). A task whose folder <WORKSPACE>/*/<KEY>/ exists outside SPRINT_DIR (an earlier sprint, or tasks/) is carried over: its session and documents stay there; create no folder or session for it. Missing = unfinished, not carried over and no session. Finished tasks and tasks that left the sprint are never touched.

4. Prepare folders

mkdir -p SPRINT_DIR/<KEY> for every unfinished task. Never delete or overwrite anything in an existing task folder.

5. Open the missing task sessions

For each missing task, in key order:

• Title: <KEY> + the summary with leading [...] tags and spaces removed; over 60 characters: cut to 59 and append ….
• Fill references/task-session-prompt.md and call mcp__ccd_session__start_session with initiation: "user_asked", context: "fresh", cwd: <task dir>, use_worktree: false, title, prompt = filled Prompt section, background = filled Background section. Do not pass permission_mode or model.
• Do not move task sessions: they nest under this session and follow it.

Before the batch, write a bold title line and one sentence.

6. Write summary.md

Write SPRINT_DIR/summary.md in DOC_LANG following references/summary-format.md.

7. Report

Send report.done. If sessions were opened, send report.ready once when each has finished its first turn, and stop.