next-feature - SKILL.md Agent Skill

name: next-feature description: Select the next feature to implement from agent-work/features.yaml. metadata: thinkingLevel: minimal

Assume SKILLS_ROOT is set per AGENTS.md before running helper commands.

1. Review State

Run git log --oneline -10 to understand recent work
Run $SKILLS_ROOT/_lib/features_yaml.sh next to find the next actionable feature
Read docs/STRUCTURE.md only if feature context is unclear

2. Select Feature

Priority order: in_progress first (resume active work), then pending features whose depends_on are all done, ranked by priority (1 > 2 > 3), then created_at, then ID.

If nothing is ready, report blocked items and their unmet deps — do not guess or auto-resolve cycles.

Report active work and a few next-ready options. For multi-repo sessions, report per repo.

Default helper output has three sections:

IN PROGRESS
- [id] (priority [n]): [description]

READY OPTIONS
1. [id] (priority [n], deps: [list or "none"])
   [description]

RECOMMENDED NEXT
[id]
Suggested plan file: agent-work/plans/[id].md

Rules:

IN PROGRESS lists all active tracked features in the same order used for recommendation.
READY OPTIONS lists up to the top few ready pending features by the same priority and tie-break rules as selection.
RECOMMENDED NEXT remains the single canonical next item.
If nothing is actionable, report the no-ready situation and include the first few blocked items with unmet dependencies instead of inventing a recommendation.

Do not modify agent-work/features.yaml. Status changes happen in execute/commit.

Output

When an actionable feature is recommended, end with READY FOR PRIME. If nothing is actionable, report the blocked/no-ready state instead.