Feature Mode

Document a new feature within an existing product. Generates 3-5 focused documents without requiring the full 4-phase pipeline.

Usage

/feature                         -> Start or resume a feature workflow
/feature [description]           -> Start with feature description
/feature status [slug]           -> Show current feature state and next recommended step
/feature ideate [slug|context]   -> Run feature-scoped ideation
/feature solution [slug|context] -> Run feature-scoped solution discovery
/feature brief [slug|context]    -> Generate or refine the feature brief
/feature stories [slug]          -> Generate scoped user stories
/feature delta [slug]            -> Generate the design delta
/feature impl [slug]             -> Generate the implementation plan

Agent Dispatch

Feature mode dispatches the planning-agent (Opus) for solution discovery, feature brief, and user stories, and the architecture-agent (Sonnet) for design delta and implementation plan. The orchestrator determines which agent to spawn based on the current step.

Workflow

1. Detect or confirm Feature mode
2. Resolve or create a canonical feature slug
3. Create or read the state hub:
   • docs/ets/projects/{project-slug}/features/{feature-slug}/feature-status.md
   • Run python3 .claude/hooks/state-manager.py ensure-feature {project-slug} {feature-slug} none
4. Read execution_adapter from docs/ets/projects/{project-slug}/state/project-status.yaml
5. Smart resume from the next incomplete step:
   • ideate -> solution -> brief -> stories -> delta -> impl
6. Update feature-status.md after each completed step:
   • current_step
   • last_completed_step
   • next_recommended_step
   • linked_docs
   • traceability
7. Save all feature-scoped artifacts under the same feature folder

Output

docs/ets/projects/{project-slug}/features/{feature-slug}/
├── feature-status.md           # Canonical state hub for this feature
├── solution-discovery.md       # Solution options, four-risk matrix, recommendation
├── solution-experiments.yaml   # Optional experiments (EXP-#)
├── feature-brief.md            # Problem, scope, personas, acceptance criteria
├── user-stories.md             # US-# with Given/When/Then (feature-scoped)
├── design-delta.md             # Only what changes: endpoints, DB migrations, UI
└── impl-plan.md                # impl-# tasks, T-shirt sizing, sprint assignment

Dependencies

• opportunity-pack.md — BLOCKS (generated by /ideate, mandatory)
• coverage-matrix.yaml — BLOCKS (generated by /ideate, mandatory)
• features/{feature-slug}/solution-discovery.md — BLOCKS (generated by /feature solution, mandatory)
• features/{feature-slug}/feature-status.md — Canonical state hub for resume/status operations
• project-context.md — ENRICHES (read if exists, proceed without)
• product-vision.md — ENRICHES (read if exists, proceed without)

Validation

• Thresholds lowered: US-# ≥1, acceptance criteria ≥1 per story
• Coverage matrix required before the feature brief, but scoped thresholds are lower
• Feature state must stay coherent:
  • feature-status.md exists
  • execution_adapter is inherited from project-status
  • next_recommended_step points to the next incomplete public subcommand
• No quality gates (optional with /validate)

Auto-Escalation

If scope grows beyond >5 user stories or >3 feature-specs, the orchestrator suggests switching to Product mode.

Traceability

ACT/JTBD/JOUR/UC/EDGE -> SOL-# -> FB-# → US-# → impl-# (simplified chain, no BO-# or PRD-F-# required)

$ARGUMENTS