plan

name	plan
description	Use when starting any non-trivial task - classifies task level (L0-L3), runs appropriate planning agents, produces intent document and execution plan

Plan

Classify the task, dispatch planning agents, produce an immutable intent + feature list.

Core principle: every plan step names concrete files and changes — no placeholders.

When to Use

• User runs /plan <task>
• Starting any new task that requires implementation
• After /discover confirms requirements

Permission

Directory	Access
decisions	R+W
progress	RW
solutions	R
reviews	R

Routing

This skill invokes the dispatcher; process is authoritative in code, not prose.

• Behavior: src/commands/plan.ts (runPlan)
• Classifier: src/dispatcher/agents/classifier-level.ts
• Planner cluster: planner.eng (L1+), +planner.ceo +researcher.history (L2+), +planner.adversarial (L3)
• Manifests: contracts/sgc-capabilities.yaml
• Invariants: §2 intent immutable · §4 L3 requires --signed-by + interactive yes · §11 classifier rationale specific
• Levels, permissions, escalation: plugins/sgc/CLAUDE.md

Execution

When this skill is invoked, dispatch to the sgc CLI:

bun src/sgc.ts plan "$ARGUMENTS"

For L3 tasks, the CLI will prompt for --signed-by and require interactive yes. To override auto-classified level: --level L0|L1|L2|L3 (upgrade-only).

Delegation hint

sgc plan produces intent.md + feature-list.md. For deep implementation planning:

• sp:writing-plans — task-by-task execution plan
• sp:brainstorming — clarify ambiguous scope before planning

sgc owns classification + intent; sp owns the deep plan body.