feat: close the loop — offer to create tracked follow-ups from a story's loose ends

[atamanvega]

Motivation

Today, when a story finishes, the kit describes loose ends but doesn't track them: create-pr puts them in the PR body ("Out of scope / follow-ups") and issue-update repeats them in the tracker comment. They live as prose and get lost.

Now that the kit can create tickets (the plan-backlog skill + the tracker write adapters), we can close the loop: at story completion, offer to create the loose ends as tracked follow-up work items in the configured tracker — approval-gated, linked back to the original story.

Full loop becomes: idea → backlog → ticket → PR → follow-ups tracked.

Where it hooks in

• The coding-agent orchestrator, as a final step after create-pr + issue-update: collect the follow-ups surfaced during the story and ask "Create these N follow-ups as tickets? [approve / edit / skip]".
• The portable work-story playbook (codex/skills/work-story) gets the same final step.
• fix-pr too (review can surface deferred items — it already decides fix / defer-to-issue / discard; the "defer" bucket should become a real ticket).

What it does

1. Gather the real loose ends from the story (out-of-scope notes, deferred review findings, TODOs the work deliberately left) — never invent them.
2. Propose each as a work item (default Task, or User Story when it's a user-facing increment), with a one-line description + why, and link it to the source story/epic for traceability.
3. Approval gate: show the list; create nothing until the user approves (they can edit/drop items).
4. Create via the tracker write adapters (reuse plan-backlog's creation: Jira / Linear / GitHub / Azure), verify writes by read-back where the CLI can silently no-op.
5. Report created items with keys/URLs; note them in the PR/tracker comment as "Follow-ups tracked: ".

Guardrails

• Never auto-create; approval-gated.
• Only track genuine loose ends from the work — no fabricated backlog.
• Always link to the source story for traceability.
• Report exactly what was created.

Open questions

• New small skill (e.g. follow-ups / spin-off) invoked by the orchestrator, vs extending issue-update (which already handles tracker writes + the follow-ups comment)?
• Default item type: Task vs User Story (or ask)?
• Link semantics per tracker (Jira "relates to"/sub-task; Linear sub-issue/relation; GitHub task-list/"tracked by"; Azure related link).

Proposed v1

• A follow-ups skill that takes the loose ends + the source story key, proposes items, gates, and creates them linked to the source (Jira + GitHub first, then Linear + Azure — the model covers all four).
• Wire the offer into coding-agent's final step and the work-story playbook (offer, never force).

Design captured — cc @ophiocus. v2-adjacent to #64 (reuses its creation adapters).

[atamanvega]

Implemented in #68. The follow-ups skill turns a finished story's genuine loose ends (out-of-scope notes, deferred pr-review/fix-pr findings, TODOs) into tracked work items linked to the source story — approval-gated, no fabrication, via the Jira/Linear/GitHub/Azure adapters. Wired as an offered step 8 in coding-agent and the portable work-story playbook. Full loop now: idea → backlog → ticket → PR → follow-ups tracked. Closing.