feat(flow-code): add opt-in --interactive flag to /flow-code:plan

Routes through /flow-code:interview BEFORE Context Analysis when flag
is passed. Opt-in only — no auto-trigger heuristic, no --no-interview
flag — preserves the zero-interaction default contract (CLAUDE.md:99).

Task: fn-20-abf-borrowed-enhancements-archetypes.5

Author: z23cc

skills/flow-code-interview/SKILL.md

@@ -69,9 +69,14 @@ FLOWCTL="${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/bin/flowctl"
    - Read spec: `$FLOWCTL cat <id>`
    - Also get epic context: `$FLOWCTL cat <epic-id>`
 
-3. **File path**: anything else with a path-like structure or .md extension
+3. **File path**: anything with a path-like structure or `.md` extension AND the file exists on disk
    - Read file contents
-   - If file doesn't exist, ask user to provide valid path
+   - If file doesn't exist, fall through to raw-text mode (below)
+
+4. **Raw text (plan-integration mode)**: first arg does NOT match a Flow ID pattern AND is NOT an existing file path
+   - Treat entire `$ARGUMENTS` as the raw request text to refine
+   - Conduct a focused refinement interview (see **Plan-integration mode** below)
+   - Used by `/flow-code:plan --interactive` to refine a vague request before planning
 
 ## Interview Process
 
@@ -95,6 +100,32 @@ Options: a) PostgreSQL b) SQLite c) MongoDB
 
 Read [questions.md](questions.md) for all question categories and interview guidelines.
 
+## Plan-integration mode (raw-text input)
+
+When invoked with raw request text (input type 4 above) — typically from `/flow-code:plan --interactive`:
+
+- **Hard cap: 12 questions max.** This mode is a focused pre-plan refinement, not the full 40+ question spec interview. Prioritize questions that disambiguate scope, uncover missing acceptance criteria, and surface edge cases.
+- Group 2-4 related questions per `AskUserQuestion` call (same pattern as standard mode).
+- **Do NOT write to `.flow/`.** Do not create an epic, do not call `flowctl epic create`.
+- **Output contract**: emit refined-spec markdown to stdout with exactly these four sections, then return control to the caller:
+
+  ```markdown
+  ## Problem
+  <concise problem statement distilled from raw input + answers>
+
+  ## Scope
+  <what's in / what's out, key decisions made during interview>
+
+  ## Acceptance
+  - [ ] <testable criterion 1>
+  - [ ] <testable criterion 2>
+
+  ## Open Questions
+  <unresolved items for planning research, or "None" if fully clarified>
+  ```
+
+The caller (e.g. `/flow-code:plan`) uses this refined markdown as the effective request text for its own Context Analysis and scout research.
+
 ## NOT in scope (defer to /flow-code:plan)
 
 - Research scouts (codebase analysis)

skills/flow-code-plan/SKILL.md

@@ -102,6 +102,7 @@ Research: <repo-scout|context-scout> | Depth: <short|standard|deep> | Review: <r
 
 These flags override the corresponding AI decision without entering the analysis flow:
 - `--research=rp|grep`, `--depth=short|standard|deep`, `--review=rp|codex|export|none`, `--plan-only`
+- `--interactive` — **opt-in** interview refinement. Before Context Analysis, invoke `/flow-code:interview` with the raw request text. The interview returns refined-spec markdown (Problem / Scope / Acceptance / Open Questions). Use that refined text as the effective request for Context Analysis and Step 1. When this flag is NOT passed, the plan flow is unchanged and the zero-interaction default (CLAUDE.md:99) is preserved. There is intentionally no auto-trigger heuristic and no `--no-interview` flag — interview is opt-in only.
 
 Proceed to Step 1 immediately.
 

skills/flow-code-plan/steps.md

@@ -46,6 +46,8 @@ FLOWCTL="${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/bin/flowctl"
 $FLOWCTL init --json
 ```
 
+> **Note — opt-in interactive refinement:** If the user passed `--interactive`, BEFORE running Step 0 (Context Analysis in SKILL.md), invoke `/flow-code:interview` with the raw request text. The interview returns refined-spec markdown with Problem / Scope / Acceptance / Open Questions sections; use that refined text as the effective request for Context Analysis and all subsequent steps. Without the flag, skip this entirely — Step 0.5 below remains an automated internal brainstorm and is **not** interactive. Do not add any auto-trigger heuristic (length, punctuation, verb detection); interview must be opt-in only to preserve the zero-interaction contract (CLAUDE.md:99).
+
 ## Step 0.5: Clarity Check (auto — no human input)
 
 **Clear?** (specific behavior, bug with repro, existing pattern, has acceptance criteria) → skip to Step 1.