spec

You are executing the Spec phase (1/6) of the CocoBrew lifecycle. Your objective is to capture complete project requirements through structured dialogue.

Before proceeding, verify that .cocoplus/ exists in the current directory. If it does not, output: "CocoPlus not initialized in this directory. Run $pod init to begin." Then stop.

Check .cocoplus/lifecycle/meta.json. If current_phase is not not_started and not spec:

If current_phase is build, test, review, or shipped: Output:

⚠️  WARNING: Current phase is [phase]. Re-entering Spec at this stage will invalidate downstream artifacts.

Active git worktrees from CocoHarvest (agent/stage-*) will become stale and out of sync with the new specification. You must clean them up manually with:
  git worktree remove --force agent/stage-<name>
  git worktree prune

To proceed anyway, re-run $spec with the --force flag: $spec --force

Then stop — unless the --force flag was provided in the invocation arguments. If --force was provided, output "Proceeding with forced Spec re-entry. Existing downstream artifacts will be invalidated." and continue.

Otherwise (phase is plan): Output "Current phase is [phase]. The Spec phase can only be entered from the beginning or re-entered to update requirements. Proceed? (yes/no)" If no: stop.

Requirements Capture Dialogue

Before the dialogue starts:

1. Check .cocoplus/lifecycle/bloom.md.
   • If it exists, read it first and summarize the Core Capability as the anchor for the specification.
   • If it does not exist and .cocoplus/lifecycle/meta.json does not contain "bloom_waived": true, output: "No working-backwards document found — consider running $bloom before specifying. Run $bloom --skip to suppress this message." Then continue normally.
2. Classify scope natively from the initial task description if one was provided, or from the first goal answer once available.
   • Count affected files/objects, destructive or billing-significant operations, external dependencies, EHRB-sensitive categories, and evaluation burden.
   • --quick forces Quick Flow. --full forces Full Flow.
   • If not forced, choose quick only when affected files/objects are <= 3, no EHRB category is present, acceptance criteria are concrete, and no parallel workstream is needed. Otherwise choose full.
   • Record the final "flow_type": "quick" | "full" in .cocoplus/lifecycle/meta.json.
   • If Quick Flow is selected, the lifecycle may skip $plan after spec capture and proceed to a single structured build artifact at .cocoplus/lifecycle/quick-build.md.

Ask each question in sequence. Wait for the developer's response before proceeding to the next question. Do not batch questions.

Question 1: What is the primary goal of this project? (Be specific — what problem does it solve and for whom?)

Question 2: What are the success criteria? (How will you know this project succeeded? List 2-5 measurable outcomes.)

Question 3: What is explicitly out of scope for this work? (List anything the project will not attempt in this phase.)

Question 4: Which existing Snowflake objects are involved? (List tables, views, stages, functions, procedures, or schemas already in play. If none, say "None".)

Question 5: Who are the target users? (Who will use or benefit from this output? If unknown, say "TBD".)

Question 6: What is the target timeline? (If unknown, say "TBD".)