feature-implement-complex

Purpose

Execute high-complexity features with full design, planning, and implementation phases.

Inputs

• feature_id
• feature_name
• test_steps
• complexity_score (26-40)

Outputs

• Design exploration result
• Implementation plan path
• Execution completed status
• Feature AI metrics with selected_model = opus

State Read/Write

Read:

• docs/progress-tracker/state/progress.json
• docs/progress-tracker/architecture/architecture.md (if present)

Write:

• docs/progress-tracker/state/progress.json via progress_manager.py

Execution Context Requirements

CRITICAL: All plugins/progress-tracker/prog commands MUST be executed from the project root directory.

If not already in the project root, first run:

cd <project-root>

The prog tool uses relative paths and requires being in the correct project directory to resolve files correctly.

Steps

1. Validate complexity bucket is complex.
2. Display complex-mode banner and rationale.
3. Run unified preflight and follow decision before design/execution:

plugins/progress-tracker/prog git-auto-preflight --json

• If decision=REQUIRE_WORKTREE, run:

Skill("using-git-worktrees", args="Set up isolated workspace for feature-<id>")

• If decision=DELEGATE_GIT_AUTO, run:

Skill("progress-tracker:git-auto", args="Resolve workspace/git preflight blockers for feature-<id>")

• If decision=ALLOW_IN_PLACE, continue in current workspace.

4. Run design phase:

Skill("brainstorming", args="<feature_name>: architecture and approach")

If docs/progress-tracker/architecture/architecture.md exists, include Execution Constraints (CONSTRAINT-*) in brainstorming context.

5. Update workflow to design complete:

plugins/progress-tracker/prog set-workflow-state \
  --phase "design_complete" \
  --next-action "planning"

6. Run planning phase:

Skill("writing-plans", args="<feature_name>: create implementation plan\nArchitecture constraints:\n- <CONSTRAINT-...>\nPlan path policy: must output under docs/plans/YYYY-MM-DD-<slug>.md (e.g. docs/plans/2026-03-03-feature-name.md)")

7. Update workflow to planning complete with plan path:

plugins/progress-tracker/prog set-workflow-state \
  --phase "planning_complete" \
  --plan-path "<returned_plan_path>" \
  --next-action "execution"

8. Run execution phase:

Skill("subagent-driven-development", args="plan:<returned_plan_path>")

9. Run final review + verification gates:

Skill("requesting-code-review", args="Review complex feature implementation: <feature_name>")
Skill("verification-before-completion", args="Verify complex feature evidence for <feature_name>")

10. Mark workflow as execution complete and save AI metrics:

plugins/progress-tracker/prog set-workflow-state \
  --phase "execution_complete" \
  --next-action "verify_and_complete"

plugins/progress-tracker/prog set-feature-ai-metrics <feature_id> \
  --complexity-score <score> \
  --selected-model opus \
  --workflow-path full_design_plan_execute

11. Ask user to run /prog done.

Compatibility rule:

• Do not perform branch finalization here.
• This skill prepares implementation for /prog done, which owns feature completion.

Plan Path Policy (Required)

set-workflow-state --plan-path must only store docs/plans/*.md. If returned path is invalid or missing, regenerate the plan before execution.

Failure Modes

• If any phase fails, preserve the latest workflow state and return clear recovery guidance.
• Do not mark feature complete here.

Commands

• plugins/progress-tracker/prog set-workflow-state ...
• plugins/progress-tracker/prog set-feature-ai-metrics ...

Examples

• Auth system refactor (OAuth2 + JWT)
• Multi-module integration
• Core architecture redesign