forge-sprint — a sprint view over the task DAG

/forge:sprint slices the project's task DAG into bounded sprints: plan commits the next ready tasks to a sprint file; review reports what got done vs. what carries over. It is not a parallel tracker — it references T-IDs only, and the task DAG (pipeline/05-plan/task-dag.md) plus implementation progress (pipeline/06-implementation/progress.md) remain the single source of truth (F-044/045).

The selection is deterministic — scripts/sprint.py does it (no LLM); this skill invokes it and reports the result.

When to Use

• /forge:sprint plan — commit the next chunk of ready tasks to sprint-NN.md.
• /forge:sprint review — at sprint end, summarize done vs. carried into a review file.
• /forge:sprint list — show existing sprints and their progress.

When NOT to Use

• You just want to build a milestone end-to-end → /forge:build --milestone N.
• There is no task DAG yet → run /forge:plan (Stage 5) first.

Steps

1. Plan a sprint.

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/sprint.py plan --cwd . \
     [--size K] [--milestone N]

   It selects the next K (default 5) ready tasks in dependency order — carry-over first: any task committed to the previous sprint that is not yet done in progress.md leads the new sprint and keeps its T-ID (F-047) — then writes pipeline/05-plan/sprint-NN.md and prints the committed T-IDs to stdout. Show the user the sprint contents. If it reports "nothing to plan", everything is done — say so.
2. Work the sprint. Build its tasks the normal way (/forge:build, per-task or --milestone). Mark tasks done in progress.md as usual — the sprint file is just a view; you do not edit it by hand.
3. Review at sprint end.

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/sprint.py review --cwd . [--sprint NN]

   Defaults to the latest sprint. It writes pipeline/12-release/sprint-NN-review.md (Done / Carried / Blockers / Lessons) by cross-referencing the sprint's T-IDs against progress.md. Summarize: how many done, what carries over, any blockers. Carried tasks automatically lead the next /forge:sprint plan.
4. List anytime.

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/sprint.py list --cwd .

Notes

• Opt-in (F-048). Sprint files live under pipeline/05-plan/ and pipeline/12-release/ and are created only when you run this skill. A project that never uses sprints is unaffected.
• T-IDs are identity (F-047). Carry-over and review track tasks by T-ID, so a task that spans sprints keeps its history. Never renumber tasks to "fit" a sprint.
• The script writes only sprint files; it never mutates the DAG, progress.md, or pipeline state. It degrades cleanly (exit 1 + a message) when there is no DAG.