Writing Plans
Overview
Write implementation plans as a plan directory — a tight orchestration plan plus per-task briefing files for agents. Assume agents have zero codebase context. Document everything they need: which files to touch, complete code, testing, validation criteria. DRY. YAGNI. TDD. Frequent commits.
Announce at start: "I'm using the writing-plans skill to create the implementation plan."
Context: Can be invoked standalone or as a handoff from brainstorming. Works in the current directory or a dedicated worktree.
Save plans to: .claude/plans/YYYY-MM-DD-<feature-name>/
See references/plan-lifecycle.md for directory lifecycle (creation, execution, cleanup, stale detection).
Plan Directory Structure
.claude/plans/YYYY-MM-DD-<feature-name>/
plan.md # Orchestration plan (<200 lines)
manifest.json # Machine-readable task/wave metadata
briefings/
task-01.md # Per-task agent briefing
task-02.md
...Format Specifications
See references/formats.md for plan.md template, manifest.json schema, and briefing authoring rules.
Briefing file template: references/briefing-template.md.
Execution Handoff
After saving the plan directory, offer execution:
"Plan complete and saved to .claude/plans/<plan-id>/. How would you like to execute it?"
Options:
- Execute now — I'll dispatch agents per task with review between each (stays in this session)
- Execute in separate session — Open a new session and invoke plan-execution for batch execution
REQUIRED SUB-SKILL: Use plan-execution for either option.