plan-2-v2-clarify

Please deep think / ultrathink as this is a complex task.

plan-2-v2-clarify (soft-deprecated)

This skill is the mid-plan re-entry point for clarifications. The original "create-spec-then-interrogate-spec" two-skill flow has been collapsed into plan-1b-v3-specify-and-clarify, which front-loads questions before the spec is sketched.

Use this skill ONLY when:

• A spec already exists (created by plan-1b-v3 or the legacy plan-1b-v2)
• Plan-3 / plan-2c-workshop / plan-6 implementation surfaced new ambiguities
• The user wants to add a clarification round mid-stream without regenerating the spec

For new specs, use /plan-1b-v3-specify-and-clarify instead.

User input:

$ARGUMENTS

# Expects: path to an existing spec, or a plan slug.

Tool-capability detection

• Batched host (e.g., Claude Code's AskUserQuestion): submit the round as ONE batched call (up to 4 questions).
• Sequential host: submit one at a time, preserving the cap.

Flow

1. Resolve PLAN_DIR from the spec path provided; set FEATURE_SPEC = ${PLAN_DIR}/<slug>-spec.md.
2. Scan the existing spec for unresolved gaps:
   • [NEEDS CLARIFICATION: …] markers
   • Missing or thin Testing Strategy / Documentation Strategy / Target Domains
   • Open Questions section entries
   • Domain Review needed (Target Domains contains new/contested entries not yet reviewed)
   • Agent Harness decision missing
3. Choose up to 4 highest-impact questions from the gap list. Skip questions already answered in earlier ### Session blocks.
4. Submit as ONE batched prompt (batched host) or sequentially (fallback). Cap = 4.
5. Append answers to ## Clarifications → ### Session YYYY-MM-DD (new block — do not modify earlier sessions).
6. Update affected spec sections immediately (Target Domains, Testing Strategy, Documentation Strategy, ACs, Risks, etc.). Save.
7. Emit a one-line coverage summary: "Resolved N/M open gaps. Remaining: …".

Question catalogue (draw from these, only if relevant)

See plan-1b-v3-specify-and-clarify/SKILL.md → Standard Questions for the full answer tables. Categories:

• Workflow Mode — only if Mode is unset (rare for re-entry)
• Testing Strategy — only if ## Testing Strategy is missing
• Mock Usage — only if testing exists but mock policy is unset
• Documentation Strategy — only if ## Documentation Strategy is missing
• Domain Review — only if Target Domains has unreviewed NEW/contested entries
• Agent Harness Readiness — only if harness decision not recorded
• Topic-specific — drawn from [NEEDS CLARIFICATION] markers (data model, FRs, NFRs, integrations, edge cases, terminology)

Gates

• Spec file must exist; ERROR if not found
• At least one unresolved gap identified; if none, exit silently with "No clarifications needed."
• Cap of 4 questions per invocation (run again for further rounds)
• No critical [NEEDS CLARIFICATION] markers remaining after the session (or surface the survivors in the coverage summary)

Output

Updated SPEC_FILE with a new ### Session YYYY-MM-DD block and any section updates the answers triggered.

Next step: Run **/plan-3-v3-architect** (if architect hasn't run) or re-run the downstream skill that surfaced the ambiguity.