fix(pipeline): five quality gaps — domain skills, brainstorm→plan, planSync, investigation, memory (v0.1.36)

1. Worker domain skills: expand from frontend-only to all 6 domains with
   per-domain focus table and fallback guidelines
2. Brainstorm→Plan: make requirements doc consumption mandatory with
   traceability (MUST read, MUST reference in epic spec)
3. planSync: update docs to reflect default-enabled state (was documented
   as disabled but default config already had enabled:true)
4. Worker Phase 3 (Investigation): include in S/M task sequences — all
   tasks need context before coding, not just L tasks
5. Memory system: change default from false to true so cross-epic
   knowledge accumulates automatically

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: z23cc

agents/worker.md

@@ -247,12 +247,23 @@ Parse the spec carefully. Identify:
 - **Domain** (from task JSON `domain` field): if set (frontend/backend/architecture/testing/docs/ops), focus your approach accordingly — e.g., backend tasks prioritize API/DB, frontend tasks prioritize UI/UX
 
 **Domain-specific skill loading:**
-If task domain is `frontend`, you MUST Read the skill file and follow it:
+Based on the task `domain` field, you MUST Read and follow the corresponding skill file. This is a quality gate — not optional.
+
+| Domain | Skill file to load | Focus |
+|--------|-------------------|-------|
+| `frontend` | `skills/flow-code-frontend-ui/SKILL.md` | Component architecture, design system, accessibility, AI aesthetic avoidance |
+| `backend` | Apply `flow-code-api-design` patterns (if skill exists) | API design, DB queries, error handling, input validation |
+| `testing` | Apply `flow-code-debug` patterns | Test coverage, edge cases, regression guards |
+| `docs` | Follow project's doc conventions | Accuracy, completeness, cross-references |
+| `architecture` | Apply `flow-code-api-design` patterns | Module boundaries, dependency direction, interface stability |
+| `ops` | Focus on idempotency, rollback safety, monitoring | CI/CD, infra, deploy scripts |
+
 ```bash
-# Load frontend UI engineering skill (component architecture, design system, accessibility, AI aesthetic avoidance)
-cat "${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}/skills/flow-code-frontend-ui/SKILL.md"
+# Example: load frontend skill
+PLUGIN_ROOT="${DROID_PLUGIN_ROOT:-${CLAUDE_PLUGIN_ROOT}}"
+cat "$PLUGIN_ROOT/skills/flow-code-frontend-ui/SKILL.md"
 ```
-Apply all phases from the skill during implementation. This is not optional — it is a quality gate.
+If the skill file does not exist for a domain, apply the domain focus guidelines from the table above.
 
 **Baseline check:**
 ```bash
@@ -277,7 +288,7 @@ Save `GIT_BASELINE_REV` — you'll use it in Phase 10 to generate workspace chan
 <!-- section:core -->
 ## Phase 3: Pre-implementation Investigation
 
-**If the task spec contains `## Investigation targets` with content, execute this phase. Otherwise skip to Phase 4/5.**
+**Always execute this phase** — even S/M tasks need context before coding. If the task spec contains `## Investigation targets`, follow them. If not, do a lightweight scan: read the files listed in `**Files:**` and check for 2-3 related patterns via Grep. Skip only if the task is a trivial one-line config change with no dependencies.
 
 ### Step 0: RP-powered deep context (if RP_CONTEXT != none)
 

bin/flowctl

@@ -63,14 +63,15 @@ elif ls .flow/specs/*-requirements.md 1>/dev/null 2>&1; then
 fi
 ```
 
-**If a brainstorm requirements doc exists**, read it and use as enriched context:
-- Extract `## Problem`, `## Requirements`, `## Constraints`, `## Non-Goals` sections
-- Use `## Chosen Approach` to guide scout research direction
+**If a brainstorm requirements doc exists**, you MUST read it and use as enriched context (this is NOT optional — brainstorm output is a first-class input to planning):
+- Extract `## Problem`, `## Requirements`, `## Constraints`, `## Non-Goals` sections — these override any guesses
+- Use `## Chosen Approach` to guide scout research direction and as the primary approach (do not re-derive from scratch)
 - Use `## Evidence` file references as starting points for repo-scout
-- Use `## Open Questions` as specific research targets for scouts
+- Use `## Open Questions` as specific research targets for scouts — each MUST be resolved or explicitly carried as a gap
 - Pass `## Self-Interview Trace` (if present) as additional context for deep RP analysis
+- Reference the requirements doc in the epic spec: `**Source:** .flow/specs/<slug>-requirements.md`
 
-This means `/flow-code:brainstorm` → `/flow-code:plan` flows seamlessly: brainstorm output directly enriches plan research instead of being orphaned.
+This means `/flow-code:brainstorm` → `/flow-code:plan` flows seamlessly: brainstorm output directly enriches plan research instead of being orphaned. The plan MUST be traceable back to the requirements doc.
 
 **If no brainstorm doc exists**, proceed normally — Step 2 does its own mini clarity check.
 

codex/skills/flow-code-plan/steps.md

@@ -392,7 +392,7 @@ Check if plan-sync should run:
 $FLOWCTL config get planSync.enabled --json
 ```
 
-Skip unless planSync.enabled is explicitly `true` (null/false/missing = skip).
+Plan-sync is **enabled by default** (default config: `planSync.enabled: true`). Skip only if explicitly set to `false`.
 
 Get remaining tasks (todo status = not started yet):
 
@@ -429,11 +429,11 @@ CROSS_EPIC: $CROSS_EPIC
 Follow your phases in plan-sync.md exactly.
 ```
 
-**Cross-epic sync** (disabled by default): When `planSync.crossEpic` is `true`, plan-sync also checks other open epics for stale references to the completed task's APIs. Enable via:
+**Cross-epic sync** (enabled by default): When `planSync.crossEpic` is `true`, plan-sync also checks other open epics for stale references to the completed task's APIs. Disable via:
 ```bash
-$FLOWCTL config set planSync.crossEpic true
+$FLOWCTL config set planSync.crossEpic false
 ```
-Use for multi-epic projects where epics share APIs. Disable for single-epic work (default) to save tokens.
+Disable for single-epic work to save tokens. Keep enabled (default) for multi-epic projects where epics share APIs.
 
 Plan-sync returns a summary. **Check the result before advancing:**
 

codex/skills/flow-code-work/phases.md

@@ -1,6 +1,6 @@
 [package]
 name = "flowctl-cli"
-version = "0.1.35"
+version = "0.1.36"
 description = "CLI entry point for flowctl"
 edition.workspace = true
 rust-version.workspace = true

flowctl/crates/flowctl-cli/Cargo.toml

@@ -81,16 +81,17 @@ const PHASE_DEFS: &[PhaseDef] = &[
 const CANONICAL_ORDER: &[&str] = &["1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12"];
 
 /// Default phase sequence (Worktree + Teams, always includes Phase 1).
+/// Phase 3 (investigation) now included by default — even S/M tasks need context.
 /// Phase 9 is inserted before 10 when `outputs.enabled` is true (default).
-const PHASE_SEQ_DEFAULT: &[&str] = &["1", "2", "5", "6", "7", "10", "11", "12"];
-const PHASE_SEQ_TDD: &[&str]    = &["1", "2", "4", "5", "6", "7", "10", "11", "12"];
-const PHASE_SEQ_REVIEW: &[&str] = &["1", "2", "5", "6", "7", "8", "10", "11", "12"];
+const PHASE_SEQ_DEFAULT: &[&str] = &["1", "2", "3", "5", "6", "7", "10", "11", "12"];
+const PHASE_SEQ_TDD: &[&str]    = &["1", "2", "3", "4", "5", "6", "7", "10", "11", "12"];
+const PHASE_SEQ_REVIEW: &[&str] = &["1", "2", "3", "5", "6", "7", "8", "10", "11", "12"];
 
 /// Size-based phase sequences.
-/// S: fast path — skip investigation, outputs, memory. Keep guard (phase 6).
-const PHASE_SEQ_SMALL: &[&str]  = &["1", "2", "5", "6", "7", "10", "12"];
-/// L: thorough path — all 11 defined phases (adds investigation, TDD, review, outputs, memory).
-const PHASE_SEQ_LARGE: &[&str]  = &["1", "2", "4", "5", "6", "7", "8", "9", "10", "11", "12"];
+/// S: fast path — includes lightweight investigation (phase 3), skips outputs/memory.
+const PHASE_SEQ_SMALL: &[&str]  = &["1", "2", "3", "5", "6", "7", "10", "12"];
+/// L: thorough path — all 11 defined phases (adds TDD, review, outputs, memory).
+const PHASE_SEQ_LARGE: &[&str]  = &["1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12"];
 
 fn get_phase_def(phase_id: &str) -> Option<&'static PhaseDef> {
     PHASE_DEFS.iter().find(|p| p.id == phase_id)
@@ -99,9 +100,9 @@ fn get_phase_def(phase_id: &str) -> Option<&'static PhaseDef> {
 /// Build the phase sequence based on mode flags and task size.
 ///
 /// Size controls the base sequence:
-///   - S (small): fast path — 7 phases, skips investigation/outputs/memory
-///   - M (medium, default): standard 8-phase sequence
-///   - L (large): thorough — all 11 defined phases
+///   - S (small): fast path — 8 phases, includes investigation, skips outputs/memory
+///   - M (medium, default): standard 9-phase sequence with investigation
+///   - L (large): thorough — all 12 defined phases
 ///
 /// Additional flags (--tdd, --review) merge extra phases into the base.
 /// Config overrides (outputs.enabled, memory.enabled) apply on top.
@@ -143,7 +144,7 @@ fn build_phase_sequence(tdd: bool, review: bool, size: TaskSize) -> Vec<&'static
     if read_config_bool("outputs.enabled", true) {
         phases.insert("9");
     }
-    if read_config_bool("memory.enabled", false) {
+    if read_config_bool("memory.enabled", true) {
         phases.insert("11");
     }
 
@@ -358,8 +359,8 @@ mod tests {
     #[test]
     fn test_build_sequence_size_s() {
         let seq = build_phase_sequence(false, false, TaskSize::Small);
-        // S base always includes these core phases
-        for p in &["1", "2", "5", "6", "7", "10", "12"] {
+        // S base includes investigation (phase 3) since all tasks need context
+        for p in &["1", "2", "3", "5", "6", "7", "10", "12"] {
             assert!(seq.contains(p), "S sequence missing phase {p}");
         }
         // S should NOT include TDD or review without flags
@@ -370,8 +371,8 @@ mod tests {
     #[test]
     fn test_build_sequence_size_l() {
         let seq = build_phase_sequence(false, false, TaskSize::Large);
-        // L base includes TDD, review, outputs — all non-conditional phases
-        for p in &["1", "2", "4", "5", "6", "7", "8", "10", "12"] {
+        // L base includes TDD, review, outputs, investigation — all non-conditional phases
+        for p in &["1", "2", "3", "4", "5", "6", "7", "8", "10", "12"] {
             assert!(seq.contains(p), "L sequence missing phase {p}");
         }
         // L is strictly a superset of S
@@ -385,15 +386,16 @@ mod tests {
     fn test_size_s_with_tdd() {
         let seq = build_phase_sequence(true, false, TaskSize::Small);
         assert!(seq.contains(&"4"), "S+TDD should include phase 4");
+        assert!(seq.contains(&"3"), "S+TDD should include investigation");
         assert!(seq.contains(&"5"));
         assert!(seq.contains(&"6"));
     }
 
     #[test]
     fn test_backward_compat_no_size() {
         let seq = build_phase_sequence(false, false, TaskSize::Medium);
-        // Medium always includes the default core phases
-        for p in &["1", "2", "5", "6", "7", "10", "12"] {
+        // Medium always includes investigation + core phases
+        for p in &["1", "2", "3", "5", "6", "7", "10", "12"] {
             assert!(seq.contains(p), "M sequence missing phase {p}");
         }
         // No TDD, no review