Initial commit

examples/pantry-planner/.claude/commands/speckit.implement.md

@@ -0,0 +1,13 @@
+# /speckit.implement — execute the task list
+
+Gate: tasks.md exists for the current feature (run `python3 speckit.py status`
+to confirm); if it is missing, stop and run /speckit.tasks first.
+
+1. Read spec.md, plan.md and tasks.md of the current feature.
+2. Work through tasks.md strictly top to bottom, phase by phase: implement
+   each task test-first, and flip its checkbox to `- [x]` only after its
+   verification actually passed.
+3. Never renumber or delete tasks; append newly discovered work as new
+   checkboxes at the end of the fitting phase.
+4. When all boxes are checked, run `python3 speckit.py check` and report the
+   result to the user together with how to run the built feature.

examples/pantry-planner/.claude/commands/speckit.plan.md

@@ -0,0 +1,13 @@
+# /speckit.plan — create the implementation plan
+
+Gate: the current feature must already have a completed spec.md — read it
+first; if it is missing or full of unresolved markers, stop and say so.
+
+1. Run `python3 speckit.py plan` — it seeds `plan.md` next to the spec
+   (the command refuses to run while spec.md is missing).
+2. Fill plan.md from the spec: the concrete stack in ## Technical Context,
+   the intended source layout in ## Structure, and each significant choice
+   with rationale and rejected alternative in ## Decisions.
+3. Check every decision against memory/constitution.md; justify or simplify
+   anything that violates an article.
+4. Run `python3 speckit.py status` to confirm the phase is recorded.

examples/pantry-planner/.claude/commands/speckit.specify.md

@@ -0,0 +1,18 @@
+# /speckit.specify — create a feature specification
+
+The user gives you a feature description. Turn it into a reviewable spec:
+
+1. Run `python3 speckit.py specify "<the description>"` from the project —
+   or, when the user provides the brief as a file, pass its path instead
+   (add `--name "<short feature name>"` when the file starts with metadata).
+   This creates `specs/NNN-slug/spec.md` and makes it the current feature.
+2. Open that spec.md and replace every placeholder section with real content:
+   prioritized user stories with acceptance scenarios (## User Stories),
+   verifiable numbered requirements and key entities (## Requirements),
+   measurable technology-agnostic outcomes (## Success Criteria), and the
+   defaults you chose for open points (## Assumptions).
+3. Where you truly cannot decide without the user, leave a marker of the form
+   NEEDS CLARIFICATION plus the question in square brackets — at most 3.
+4. Finish by running `python3 speckit.py check` and resolving any findings.
+
+Do not make technology choices here — the spec says what and why, never how.

examples/pantry-planner/.claude/commands/speckit.tasks.md

@@ -0,0 +1,13 @@
+# /speckit.tasks — break the plan into executable tasks
+
+Gate: the current feature needs a filled plan.md — read spec.md and plan.md
+before writing anything; if plan.md is missing, stop and say so.
+
+1. Run `python3 speckit.py tasks` — it seeds `tasks.md` for the current
+   feature (refused while plan.md is missing).
+2. Replace the placeholder checkboxes with real tasks: dependency-ordered
+   phases under ## Phases, one checkbox per task, exact file paths in each
+   description, user-story order preserved (P1 first).
+3. Every task must be small enough to verify on its own; note which tasks
+   are independent enough to run in parallel.
+4. Run `python3 speckit.py status` to see the open-task count.