Plan

Turn an idea into work another agent can build without guessing: requirements
in the user's words, decisions with their reasons, units that each end in a
check, and tickets whose order is checked rather than assumed.

Done when: the artifact the mode produces exists at the path in
../../references/artifacts.md and passes that mode's done check.

Modes

Pick the mode the request describes. With no mode word and nothing in the
request to decide it, use plan. Read only the reference for the mode you run.

How it works

1. Ground. Read memory/ when it exists (INDEX.md, current-state.md,
   relevant decisions), the code the work touches, and the lessons that match:

   node "${CLAUDE_SKILL_DIR}/../../scripts/lessons.mjs" --query "<the feature in a few words>"

2. Run the mode from its reference. Ask the user only what the repository
   cannot answer and what would change the plan, with your recommended answer
   attached. A small change gets its plan in the reply, not a file.

3. Write the artifact in the shape ../../references/artifacts.md gives.
   One plan file grows from brainstorm to plan; edit it in place.

4. Check it against the mode's done check before handing it on.

The habits in ../../references/principles.md apply throughout.

Works with

• Reads: the memory tree, lessons, a prototype's verdict, an issue or spec.
• Hands off: the plan to build, review (the spec lens), and verify;
  tickets to build and to project-memory next actions. To stress-test a
  plan against memory, run /project-memory grill <plan>.
• Alone: works in any repository. Without memory, ground in the code and
  Git history.

Paths in this skill and its references are relative to this skill's directory
(${CLAUDE_SKILL_DIR}). Outside Claude Code, that is the directory holding this
file, and $0 is the first word after the skill name.