cfsa-antigravity 2.13.3 → 2.14.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cfsa-antigravity",
-  "version": "2.13.3",
+  "version": "2.14.0",
   "description": "CFSA Pipeline — Constraint-First Specification Architecture for AI agents. Production-grade from line one.",
   "scripts": {
     "changeset": "changeset",

package/template/.agent/kit-sync.md

@@ -1,6 +1,6 @@
 # Kit Sync State
 
 upstream: https://github.com/RepairYourTech/cfsa-antigravity
-last_synced_commit: 474ea4346ac09791e0f6bc7a5ed771de79ee93f2
-last_synced_at: 2026-03-20T14:04:21Z
-kit_version: 2.13.3
+last_synced_commit: 89e129286cdbd5e33d6617d81cde4450f2dac77e
+last_synced_at: 2026-03-20T17:29:41Z
+kit_version: 2.14.0

package/template/.agent/skills/prd-templates/references/be-spec-template.md

@@ -9,6 +9,12 @@ Use this template when writing BE specs to `docs/plans/be/[NN-feature-name].md`.
 > **Deep Dives**: [links to consumed deep dives, if any]
 > **Status**: Draft | Review | Complete
 
+## Split Group _(include only if this spec resulted from a split)_
+
+> **Split origin**: [IA shard that was split, e.g., `09-messaging`]
+> **Companion specs**: [sibling BE spec files, e.g., `09b-agent-flow-api.md`]
+> **Shared entities**: [data models shared across companions, e.g., `Message`, `Thread`]
+
 ## IA Source Map
 
 [Which IA shard sections, deep dives, and cross-shard references

package/template/.agent/skills/prd-templates/references/deep-ideation-loading-protocol.md

@@ -0,0 +1,114 @@
+# Deep Ideation Loading Protocol
+
+> **Purpose**: When ideation output is large (many domains, many features), the single `ideation-index.md` summary is insufficient for downstream architecture decisions. This protocol reads domain-level index files to build a structured **Domain Digest Table** that captures every feature, role assignment, and decision without requiring the agent to hold hundreds of leaf-level files in context.
+
+## When to Invoke
+
+**Always invoke** this protocol when ANY of these thresholds are met (read from the `## Progress Summary` table in `ideation-index.md`):
+
+| Metric | Threshold |
+|--------|-----------|
+| Total domains | ≥ 6 |
+| Total leaf features | ≥ 50 |
+| Total surfaces | ≥ 2 (multi-product projects) |
+
+If none of these thresholds are met, the standard `ideation-index.md` read is sufficient — skip this protocol.
+
+## Procedure
+
+### Step 1 — Enumerate domain folders
+
+Read the `## Structure Map` in `ideation-index.md`. For each row in the domain tables, record:
+- Domain number and name
+- Path to the domain folder
+- Status (`[SURFACE]`, `[BREADTH]`, `[DEEP]`, `[EXHAUSTED]`)
+- Child count (from the "Children" column)
+- Surface (for multi-product projects)
+
+This produces a **domain enumeration list** — typically 6-20 entries.
+
+### Step 2 — Read each domain index
+
+For each domain in the enumeration list, read `{domain-folder}/{domain}-index.md`. Extract:
+
+1. **Children table** — every sub-domain and feature with: number, name, type, status, depth
+2. **Role Matrix** — which personas have access to which children, and at what level (Full / Config / Read-only / Reports / None)
+3. **Decision Log** — any product decisions made during ideation for this domain
+4. **Open Questions** — deferred questions tagged for downstream stages
+
+> **Context management**: Read domain indexes one at a time. Extract the structured data, then move to the next. Do not attempt to hold all domain indexes simultaneously.
+
+### Step 3 — Read CX files for cross-domain dependencies
+
+Read `ideation-cx.md` (global cross-cuts) and each domain-level `{domain}-cx.md`. Extract:
+- Cross-domain data dependencies (which entities are shared)
+- Trigger chains (domain A event → domain B action)
+- Shared entity ownership
+
+### Step 4 — Build the Domain Digest Table
+
+Compile the extracted data into a structured table:
+
+```markdown
+## Ideation Digest
+
+> Auto-generated by the Deep Ideation Loading Protocol. Source: domain-level index files.
+> This section is the primary source for entity identification, data strategy, and phasing decisions.
+
+### Domain Summary
+
+| # | Domain | Surface | Features | Sub-domains | Depth | Must Have | Roles | Key Entities |
+|---|--------|---------|----------|-------------|-------|----------|-------|-------------|
+| 01 | {name} | {surface} | {N} | {N} | {status} | {N} | {list} | {entity hints} |
+
+### Cross-Domain Dependencies
+
+| Source Domain | Target Domain | Type | Shared Entities |
+|--------------|--------------|------|----------------|
+| {source} | {target} | data / trigger / permission | {entities} |
+
+### Ideation Decisions (from domain logs)
+
+| # | Domain | Decision | Impact |
+|---|--------|----------|--------|
+| D-{NN} | {domain} | {decision} | {what this affects downstream} |
+
+### Deferred Questions (tagged for /create-prd)
+
+| # | Domain | Question | Owner |
+|---|--------|----------|-------|
+| Q-{NN} | {domain} | {question} | /create-prd |
+```
+
+### Step 5 — Derive Key Entities column
+
+For each domain, scan the Children table for entity hints:
+- Feature names that imply data objects (e.g., "Device History" → `Device`, `RepairRecord`)
+- Role Matrix intersections that imply access-controlled resources
+- Cross-cut shared entities from CX files
+
+List the implied entities in the "Key Entities" column. These are **hints**, not schema — they inform data strategy without prescribing it.
+
+### Step 6 — Write to architecture-draft.md
+
+Write the completed Domain Digest Table to `docs/plans/architecture-draft.md` as the first section (`## Ideation Digest`). If the file doesn't exist yet, create it with this section.
+
+> **Write immediately.** Do not hold the digest in memory — write it to disk as soon as it's compiled. All subsequent create-prd shards read from `architecture-draft.md`.
+
+## Downstream Consumption
+
+| Consumer | What It Uses |
+|----------|-------------|
+| `create-prd-architecture` (Step 5: Data strategy) | Key Entities column → entity identification; Cross-Domain Dependencies → data ownership and sync |
+| `create-prd-compile` (Step 9: Phasing) | Features + Must Have counts → phase complexity sizing; Sub-domain counts → dependency ordering |
+| `create-prd-security` | Roles column → access control scope; Cross-Domain Dependencies → trust boundaries |
+
+## Anti-Patterns
+
+| ❌ Wrong | ✅ Right |
+|---------|---------|
+| Reading all 295 leaf-level feature files | Reading 8-20 domain index files |
+| Skipping this protocol because "ideation-index.md has a summary" | Invoking when thresholds are met — summaries lose detail |
+| Deriving entities from the MoSCoW bullet list | Deriving entities from the Children tables + CX dependencies |
+| Holding all domain indexes in context simultaneously | Reading one at a time, extracting structured data, moving on |
+| Writing the digest at the end of create-prd | Writing it to architecture-draft.md immediately in Step 1.5 |

package/template/.agent/skills/prd-templates/references/fe-spec-template.md

@@ -9,6 +9,12 @@ Use this template when writing FE specs to `docs/plans/fe/[NN-feature-name].md`.
 > **IA Source**: [link to IA shard, or N/A for cross-cutting]
 > **Status**: Draft | Review | Complete
 
+## Split Group _(include only if this spec resulted from a split)_
+
+> **Split origin**: [IA shard that was split, e.g., `09-messaging`]
+> **Companion specs**: [sibling FE spec files, e.g., `09b-agent-flow-ui.md`]
+> **Shared entities**: [data models shared across companions, e.g., `Message`, `Thread`]
+
 ## Source Map
 
 | FE Spec Section | Source | Section/Lines |

package/template/.agent/skills/prd-templates/references/feature-ledger-protocol.md

@@ -0,0 +1,149 @@
+# Feature Tracking Ledger Protocol
+
+> **Purpose**: Track every ideation feature through the entire pipeline — from birth in ideation to final slice assignment. Catches dropped features that per-stage reconciliation tables miss.
+
+## Ledger Location
+
+`docs/plans/feature-ledger.md`
+
+## Ledger Format
+
+```markdown
+# Feature Tracking Ledger
+
+> Auto-generated by `/ideate-validate`. Updated by each pipeline stage.
+> Each column is populated by the stage that owns it (see Column Ownership below).
+
+| Feature ID | Feature Name | Domain | MoSCoW | IA Shard | IA Status | BE Spec | BE Status | FE Spec | FE Status | Phase | Slice |
+|-----------|-------------|--------|--------|----------|-----------|---------|-----------|---------|-----------|-------|-------|
+```
+
+## Column Ownership
+
+| Column(s) | Populated By | Stage |
+|-----------|-------------|-------|
+| Feature ID, Feature Name, Domain, MoSCoW | `/ideate-validate` Step 11.5 | Discovery |
+| IA Shard, IA Status | `/decompose-architecture-validate` | Design |
+| BE Spec, BE Status | `/write-be-spec-write` | Specification |
+| FE Spec, FE Status | `/write-fe-spec-write` | Specification |
+| Phase, Slice | `/plan-phase-write` Step 2.5 | Planning |
+
+## Status Values
+
+| Status | Meaning |
+|--------|---------|
+| ✅ | Covered — spec exists and includes this feature |
+| ⚠️ deferred | Explicitly deferred with documented reason |
+| ❌ missing | Not covered — no spec addresses this feature |
+| — | Column not yet populated (upstream stage hasn't run) |
+
+## Step 1 — Ledger Creation (`/ideate-validate`)
+
+**When**: After Step 11 (compile vision), before Step 12 (request review).
+
+**Procedure**:
+
+1. Read `ideation-index.md` → Structure Map → get all domain paths
+2. For each domain, read `{domain}-index.md` → Children table
+3. For each leaf-level child (type = `feature`), create one ledger row:
+   - **Feature ID**: `{domain-name}/{feature-file-basename}` (e.g., `inventory/01-categories`)
+   - **Feature Name**: from the Children table Name column
+   - **Domain**: domain name
+   - **MoSCoW**: from `ideation-index.md` MoSCoW Summary (match by feature name), or from the feature's own priority field if available
+4. All other columns → `—` (not yet populated)
+5. Write to `docs/plans/feature-ledger.md`
+
+> **Context management**: Read domain indexes one at a time, extract features, move on. Same pattern as the deep-ideation-loading protocol.
+
+## Step 2 — IA Assignment (`/decompose-architecture-validate`)
+
+**When**: After shard boundaries are approved and shard skeletons exist.
+
+**Procedure**:
+
+1. Read `docs/plans/feature-ledger.md`
+2. For each IA shard skeleton in `docs/plans/ia/`:
+   - Read the shard's `## Sub-features` or features list
+   - For each feature, find the matching row in the ledger (match by Feature ID or Feature Name)
+   - Set **IA Shard** = shard name (e.g., `02-inventory`)
+   - Set **IA Status** = `✅`
+3. **Orphan check**: Any row where IA Shard is still `—` after processing all shards → set IA Status = `❌ missing`
+4. Write updated ledger
+
+**If orphans are found**:
+> ⚠️ **Feature ledger orphan check**: [N] features from ideation have no IA shard assignment:
+> - `{Feature ID}`: {Feature Name}
+>
+> Options:
+> 1. Add to an existing shard
+> 2. Create a new shard
+> 3. Mark as explicitly deferred (with reason)
+
+**STOP** — wait for user decision on each orphan.
+
+## Step 3 — BE Coverage (`/write-be-spec-write`)
+
+**When**: After each BE spec is written.
+
+**Procedure**:
+
+1. Read `docs/plans/feature-ledger.md`
+2. Read the newly written BE spec's endpoint list
+3. For each endpoint, identify which Feature ID(s) it serves (from the IA shard → feature mapping)
+4. For each matched Feature ID, set **BE Spec** = spec file name, **BE Status** = `✅`
+5. Write updated ledger
+
+> **Note**: Not all features produce BE endpoints (e.g., purely client-side UI features). These should have BE Status = `⚠️ deferred` with reason "client-only".
+
+## Step 4 — FE Coverage (`/write-fe-spec-write`)
+
+**When**: After each FE spec is written.
+
+**Procedure**:
+
+1. Read `docs/plans/feature-ledger.md`
+2. Read the newly written FE spec's component list
+3. For each component, identify which Feature ID(s) it serves
+4. For each matched Feature ID, set **FE Spec** = spec file name, **FE Status** = `✅`
+5. Write updated ledger
+
+## Step 5 — Slice Assignment (`/plan-phase-write`)
+
+**When**: After Step 2.5 (spec coverage verification).
+
+**Procedure**:
+
+1. Read `docs/plans/feature-ledger.md`
+2. For each slice, identify which Feature ID(s) it covers (from its spec source map)
+3. For each matched Feature ID, set **Phase** = phase number, **Slice** = slice ID
+4. Write updated ledger
+
+## Verification Gate (`/plan-phase-preflight`)
+
+**When**: During the cross-layer consistency check.
+
+**Procedure**:
+
+1. Read `docs/plans/feature-ledger.md`
+2. Count features by completeness:
+   - **✅ Complete**: All columns through FE Status are ✅ or ⚠️ deferred
+   - **⚠️ Partial**: Some columns filled, some still `—`
+   - **❌ Orphaned**: IA Status or BE Status or FE Status = `❌ missing`
+3. Report:
+
+> **Feature Ledger Verification**:
+> - ✅ Complete: [N] features
+> - ⚠️ Partial: [N] features (columns not yet populated by upstream stages)
+> - ❌ Orphaned: [N] features (dropped between stages)
+
+4. **If any ❌ orphans** → **BLOCKING GATE**: "Feature ledger shows [N] orphaned features. Resolve before planning."
+5. **If any ⚠️ partial** → **WARNING**: "Feature ledger shows [N] features with incomplete coverage. These may be from stages not yet run."
+
+## Anti-Patterns
+
+| ❌ Wrong | ✅ Right |
+|---------|---------|
+| Skipping ledger updates "because it's just bookkeeping" | Every stage updates its columns — it's the only global tracking |
+| Marking orphaned features as ⚠️ deferred without a reason | Orphans are ❌ missing until explicitly deferred with documented reasoning |
+| Regenerating the ledger from scratch at each stage | Updating existing rows — the ledger is cumulative |
+| Ignoring the orphan check at plan-phase-preflight | Blocking on orphans — they represent dropped features |

package/template/.agent/skills/prd-templates/references/shard-boundary-analysis.md

@@ -74,3 +74,30 @@ Split rationale: [why these groups are independent]
 ```
 
 **After any split**: Update `docs/plans/ia/decomposition-plan.md` with the revised table and re-run the Must Have coverage gate.
+
+**Companion spec tracking**: When writing BE or FE specs for split shards, populate the `## Split Group` section in each spec file (see `be-spec-template.md` / `fe-spec-template.md`). This enables downstream discovery of sibling specs during implementation.
+
+## Total Shard Count Guidance
+
+The decomposition should produce a shard count proportional to the ideation depth. Too few shards compress complex domains and lose detail. Too many create coordination overhead that exceeds agent context capacity.
+
+### Expected Range by Ideation Scale
+
+| Ideation Scale | Domains | Expected Shards (incl. `00-infrastructure`) | Notes |
+|----------------|---------|---------------------------------------------|-------|
+| **Small** | 3–5 | 4–8 | Most domains map 1:1 to shards |
+| **Medium** | 6–8 | 7–14 | Some multi-domain splits expected |
+| **Large** | 9–12 | 12–20 | Cross-cutting shards multiply; enforce sub-feature limits aggressively |
+| **Deep** | 13+ | 16–25 | Maximum recommended. Beyond 25 → consider surface-level decomposition or domain grouping |
+
+### Total Count Thresholds (Post-Skeleton Check)
+
+After all shard skeletons are created in Step 5, count the total:
+
+| Total Shards | Action |
+|--------------|--------|
+| **≤ 20** | ✅ Proceed |
+| **21–25** | ⚠️ Warning — "Decomposition produced [N] shards. This will require [N × 3] spec documents (IA + BE + FE) and proportional phase planning. Confirm this is the intended scope or consider grouping related domains." |
+| **> 25** | 🛑 **Hard stop** — "Decomposition produced [N] shards. This exceeds the recommended maximum for sequential agent processing. Present a domain grouping proposal that reduces shard count to ≤ 25." |
+
+> **Why 25 max**: Each shard produces ~3 spec documents (IA + BE + FE). At 25 shards, that's 75 spec documents. Beyond this, cross-layer consistency checks become unreliable and phase planning produces impractical slice counts.

package/template/.agent/workflows/create-prd-architecture.md

@@ -88,6 +88,8 @@ Write the completed decisions to `docs/plans/architecture-draft.md` under a new
 
 Read .agent/skills/database-schema-design/SKILL.md and follow its schema design methodology.
 
+**Digest-aware entity identification**: If `## Ideation Digest` exists in `docs/plans/architecture-draft.md`, use its Key Entities column and Cross-Domain Dependencies table as the primary source for entity identification in sub-steps below. Cross-reference the digest's per-domain feature lists to ensure no entity is missed. If the digest does not exist, derive entities from the architecture sections written above.
+
 Load the Databases skill(s) from the `shared` surface row per the skill loading protocol. Each sub-item must be explored to field-level depth:
 
 1. **Data placement** — What lives in the database vs cache vs object storage vs external service vs local device storage? For each entity: which service owns it and why
@@ -119,6 +121,16 @@ Write the completed cross-store consistency table to `docs/plans/architecture-dr
 
 Read `.agent/skills/prd-templates/references/data-placement-template.md` for the template structure. Create `docs/plans/data-placement-strategy.md` using the template, filling each section with the data decisions confirmed above.
 
+## Completion Gate (MANDATORY)
+
+Before reporting completion or proceeding to next shard:
+
+1. **Memory check** — Apply rule `memory-capture`. Write any patterns, decisions, or blockers from this shard to `.agent/progress/memory/`. Architecture decisions have the highest downstream impact — every confirmed decision should have a `DEC-NNN` entry. If nothing to write, confirm: "No new patterns/decisions/blockers."
+2. **Progress update** — Update `.agent/progress/` tracking files if they exist.
+3. **Session log** — Write session entry to `.agent/progress/sessions/`.
+
+---
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/create-prd-security`.

package/template/.agent/workflows/create-prd-compile.md

@@ -61,7 +61,10 @@ Write the completed `## Development Methodology` section to `docs/plans/architec
 
 Load the CI/CD skill(s) from the cross-cutting section per the skill loading protocol.
 
-Break the feature inventory from `docs/plans/ideation/ideation-index.md` (MoSCoW Summary) into dependency-ordered phases.
+Break the feature inventory into dependency-ordered phases. Use **two sources**:
+
+1. **MoSCoW priorities** from `docs/plans/ideation/ideation-index.md` (MoSCoW Summary) — determines Must/Should/Could classification
+2. **Domain Digest** from `docs/plans/architecture-draft.md` (`## Ideation Digest`) — if it exists, use its per-domain feature counts, sub-domain counts, and cross-domain dependencies to inform phase complexity sizing and dependency ordering. A domain with 15 features and 3 sub-domains requires more phase capacity than one with 4 features.
 
 > **This kit does not build MVPs.** Every phase ships production-grade code —
 > fully tested, fully specified, fully accessible. Phases exist to manage

package/template/.agent/workflows/create-prd-design-system.md

@@ -134,6 +134,16 @@ If any section is incomplete, loop back to the relevant decision step and resolv
 
 ---
 
+## Completion Gate (MANDATORY)
+
+Before reporting completion or proceeding to next shard:
+
+1. **Memory check** — Apply rule `memory-capture`. Write any patterns, decisions, or blockers from this shard to `.agent/progress/memory/`. All 7 design system decisions should be reviewed for `DEC-NNN` entries. If nothing to write, confirm: "No new patterns/decisions/blockers."
+2. **Progress update** — Update `.agent/progress/` tracking files if they exist.
+3. **Session log** — Write session entry to `.agent/progress/sessions/`.
+
+---
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/create-prd-architecture`.

package/template/.agent/workflows/create-prd-security.md

@@ -114,6 +114,16 @@ Read .agent/skills/logging-best-practices/SKILL.md and follow its Observability
 
 Write the completed `## Observability Architecture` section to `docs/plans/architecture-draft.md` immediately after user confirmation. Follow the write verification protocol (`.agent/skills/prd-templates/references/write-verification-protocol.md`).
 
+## Completion Gate (MANDATORY)
+
+Before reporting completion or proceeding to next shard:
+
+1. **Memory check** — Apply rule `memory-capture`. Write any patterns, decisions, or blockers from this shard to `.agent/progress/memory/`. Security model decisions are high-impact — every confirmed decision should have a `DEC-NNN` entry. If nothing to write, confirm: "No new patterns/decisions/blockers."
+2. **Progress update** — Update `.agent/progress/` tracking files if they exist.
+3. **Session log** — Write session entry to `.agent/progress/sessions/`.
+
+---
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/create-prd-compile`.

package/template/.agent/workflows/create-prd-stack.md

@@ -103,6 +103,16 @@ Read each installed skill's SKILL.md before proceeding. Append the confirmed dec
 
 Read `.agent/workflows/bootstrap-agents.md` and call it with `PIPELINE_STAGE=create-prd` plus only the keys just decided. Bootstrap runs after **each confirmed decision**, not in a batch at the end. At the end of all tech decisions, call bootstrap once more with `ARCHITECTURE_DOC` set to the dated filename.
 
+## Completion Gate (MANDATORY)
+
+Before reporting completion or proceeding to next shard:
+
+1. **Memory check** — Apply rule `memory-capture`. Write any patterns, decisions, or blockers from this shard to `.agent/progress/memory/`. Tech stack decisions are high-impact — every confirmed decision should have a `DEC-NNN` entry. If nothing to write, confirm: "No new patterns/decisions/blockers."
+2. **Progress update** — Update `.agent/progress/` tracking files if they exist.
+3. **Session log** — Write session entry to `.agent/progress/sessions/`.
+
+---
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/create-prd-architecture`.

package/template/.agent/workflows/create-prd.md

@@ -46,6 +46,24 @@ Read `.agent/skills/prd-templates/references/engagement-tier-protocol.md` — ea
 
 ---
 
+## 1.5. Deep ideation loading (scale-aware)
+
+Read the `## Progress Summary` table in `ideation-index.md`. Check these thresholds:
+
+| Metric | Threshold |
+|--------|-----------|
+| Total domains | ≥ 6 |
+| Total leaf features | ≥ 50 |
+| Total surfaces | ≥ 2 (multi-product projects) |
+
+**If ANY threshold is met** → read `.agent/skills/prd-templates/references/deep-ideation-loading-protocol.md` and follow its full procedure. This produces a **Domain Digest Table** written to `docs/plans/architecture-draft.md` as `## Ideation Digest`.
+
+**If no threshold is met** → skip this step. The standard `ideation-index.md` read from Step 1 is sufficient.
+
+> **Why**: For large ideation outputs (50+ features across 6+ domains), the Structure Map and MoSCoW Summary in `ideation-index.md` are insufficient for accurate architecture decisions. The digest reads every domain-level `*-index.md` to capture the full feature inventory, role coverage, and cross-domain dependencies without needing to load hundreds of leaf files.
+
+---
+
 ## 2. Load skills
 
 Read each skill SKILL.md listed in the frontmatter `skills` array.
@@ -76,6 +94,11 @@ Check `.agent/skills/` for stack-specific skills. Read `.agent/skills/find-skill
 ### Step C — Run `.agent/workflows/create-prd-security.md`
 ### Step D — Run `.agent/workflows/create-prd-compile.md`
 
+**Shard failure recovery**: If any shard (A through D) fails mid-execution:
+1. Check `docs/plans/architecture-draft.md` for the last completed section
+2. Present the current state to the user: "Shard [N] failed at [step]. Draft contains sections [list]. Resume from failure point or restart the shard?"
+3. Do NOT proceed to the next shard until the current shard completes cleanly
+
 ---
 
 ## Step E — Quality gate

package/template/.agent/workflows/decompose-architecture-structure.md

@@ -62,9 +62,16 @@ Fallback for domains not covered in the ideation domain folders is defined in th
 - Every shard in the decomposition plan has a corresponding `.md` file
 - `00-infrastructure.md` exists regardless of domain decomposition
 - No empty files (0 bytes)
+- **Total shard count** is within acceptable range — read `.agent/skills/prd-templates/references/shard-boundary-analysis.md` → "Total Count Thresholds" and apply the ≤20/21-25/>25 gate
 
 If any skeleton is missing → create it now. Do not proceed to index creation with missing skeletons.
 
+### Domain coverage check
+
+Read `docs/plans/ideation/ideation-index.md` → Domain Map. For each domain marked as having **Must Have** features:
+- Verify at least one shard skeleton references this domain
+- If a domain has no corresponding shard → **STOP**: "Ideation domain `[name]` with Must Have features has no IA shard. Add it to the decomposition plan or explain why it's excluded."
+
 ## 6. Create IA index
 
 Read `.agent/skills/prd-templates/references/decomposition-templates.md` for the **IA Index** template. Create `docs/plans/ia/index.md` using the template, populating the shards table with all shards created in Step 5.
@@ -83,6 +90,16 @@ Read `.agent/skills/prd-templates/references/decomposition-templates.md` for the
 
 For multi-surface projects, each surface's own `index.md` contains the standard three-layer table (IA/BE/FE) scoped to that surface, following the same format as the single-surface master index.
 
+## 9.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/decompose-architecture-validate`.

package/template/.agent/workflows/decompose-architecture-validate.md

@@ -70,10 +70,28 @@ Verify structural integrity:
 - [ ] BE/FE indexes exist with conventions templates
 - [ ] Multi-surface: shared shards have lower numbers; cross-surface deps point to shared/
 
+## 12.5. Update feature tracking ledger
+
+If `docs/plans/feature-ledger.md` exists, read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow **Step 2 — IA Assignment**.
+
+For each IA shard skeleton, match its features to ledger rows and populate the IA Shard and IA Status columns. Run the orphan check — any ideation feature with no IA shard assignment is flagged as `❌ missing`. Present orphans to the user for resolution before proceeding.
+
+If the ledger does not exist, skip this step (ideation was run before the ledger protocol existed).
+
 ## 13. Generate spec pipeline tracker
 
 Read `.agent/skills/session-continuity/protocols/07-spec-pipeline-generation.md` and follow the Spec Pipeline Generation Protocol.
 
+## 13.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 14. Request review and propose next steps
 
 Use `notify_user` to present: IA directory, BE index, FE index, master index, spec pipeline tracker.

package/template/.agent/workflows/ideate-discover.md

@@ -164,6 +164,26 @@ After all features are deepened, systematically identify emergent capabilities t
 
 ---
 
+## Post-Discovery Verification
+
+Before completing discovery, count and verify the ideation output:
+1. List all files in `docs/plans/ideation/` recursively
+2. Verify each domain in the ideation-index has its corresponding folder with at minimum:
+   - `{slug}-index.md` (domain index)
+   - `{slug}-cx.md` (cross-cutting concerns file)
+3. Verify `ideation-index.md` exists and has a populated Domain Map
+4. If any domain folder is missing its required files → create them now before proceeding
+
+## Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ### Next step
 
 **STOP** — do NOT proceed to any other workflow. The only valid next step is `/ideate-validate`.

package/template/.agent/workflows/ideate-validate.md

@@ -116,6 +116,16 @@ Verify every domain in `ideation-index.md` appears in `vision.md`. Nothing dropp
 
 ---
 
+## 11.5. Create feature tracking ledger
+
+Read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow **Step 1 — Ledger Creation**.
+
+Enumerate every leaf feature from the ideation fractal tree and write `docs/plans/feature-ledger.md` with Feature ID, Feature Name, Domain, and MoSCoW columns populated. All downstream columns start as `—`.
+
+> **This step is non-optional.** The ledger is the only global tracking mechanism that follows features from ideation through implementation. Without it, features can be silently dropped at stage boundaries.
+
+---
+
 ## 12. Request review
 
 ### Self-check against Ideation rubric

package/template/.agent/workflows/implement-slice-setup.md

@@ -80,6 +80,16 @@ For each acceptance criterion, trace its spec citation (e.g., `[BE §3.2]`, `[FE
 2. Read the full §section from every cited FE spec — component props, states, interactions, responsive behavior, accessibility rules
 3. Read any IA shard sections cited by the BE spec's Source Map — especially `## Edge Cases` and `## Access Control`
 
+### Companion spec discovery (split-awareness)
+
+After loading the cited specs, check each spec filename for a letter suffix (e.g., `09a-`, `09b-`). If found:
+
+1. Check the spec for a `## Split Group` section. If it lists **Companion specs**, read the `## Split Group` section from each companion to load shared entity definitions
+2. If no `## Split Group` section exists, fall back to directory scan: list all files in the same `docs/plans/be/` or `docs/plans/fe/` directory with the same number prefix (e.g., all `09*` files). Read the header block of each match to identify siblings
+3. Load any **Shared entities** into context — these data models are used across companions and may affect contracts, validation, or state management in the current slice
+
+> **Why**: A split shard divides a domain for spec-writing manageability, but the underlying data models and access patterns remain shared. Without loading companion context, the implementation may miss cross-cutting validation rules or produce incompatible data shapes.
+
 This context persists throughout the TDD cycle. The acceptance criteria define WHAT to test; the spec context defines HOW DEEP to test it.
 
 ## 1.5. Check for parallel mode

package/template/.agent/workflows/plan-phase-preflight.md

@@ -120,6 +120,16 @@ Collect all mismatches into a consistency report. If any exist → **hard stop**
 
 ---
 
+## 0.7. Feature ledger verification
+
+If `docs/plans/feature-ledger.md` exists, read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow the **Verification Gate** procedure.
+
+Count features by completeness (✅ complete, ⚠️ partial, ❌ orphaned). If any features show `❌ missing` in the IA Status, BE Status, or FE Status columns → **BLOCKING GATE**: present the orphaned features and require resolution before proceeding to slice planning.
+
+If the ledger does not exist, skip this step.
+
+---
+
 ## 0.8. Draft continuity check
 
 Check whether `docs/plans/phases/phase-N-draft.md` already exists (where N is the current phase number).

package/template/.agent/workflows/plan-phase-write.md

@@ -73,6 +73,21 @@ After all slices are identified, verify that the slices collectively cover ALL s
 
 **BLOCKING GATE**: Do NOT proceed to Step 3 until every BE endpoint and FE component is either assigned to a slice or explicitly deferred.
 
+**Feature ledger update**: If `docs/plans/feature-ledger.md` exists, read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow **Step 5 — Slice Assignment**. Map each slice to its Feature IDs and populate the Phase and Slice columns.
+
+### 2.75. Split companion cross-reference
+
+For each spec in the phase scope, check if its filename contains a letter suffix (e.g., `09a-`, `09b-`):
+
+1. If a letter suffix is found, read the spec's `## Split Group` section to identify companion specs
+2. If any companion spec is ALSO in the current phase scope, verify that slices which reference entities shared across the split are annotated with companion citations. Example:
+   - Slice "Create message thread" cites `[BE §3.2 09a-chat-api]` and uses the `Thread` entity
+   - `Thread` is listed as a shared entity with `09b-agent-flow-api`
+   - → Add companion context note: "**Companion context**: `Thread` entity shared with `09b-agent-flow-api.md` — load `§ Database Schema` from companion during implementation"
+3. If a companion spec is NOT in the current phase scope, add a note to the slice: "**Cross-phase dependency**: Companion `09b-agent-flow-api.md` is in Phase N+1 — shared entity `Thread` must remain backward-compatible"
+
+This step prevents implementation from producing incompatible data shapes across split domain boundaries.
+
 ## 3. Order by dependency
 
 Read .agent/skills/concise-planning/SKILL.md and follow its methodology.

package/template/.agent/workflows/setup-workspace-scaffold.md

@@ -23,6 +23,15 @@ Initialize the project, create the directory structure, install dependencies, an
 
 ---
 
+## 0.5. Feature ledger pre-flight
+
+If `docs/plans/feature-ledger.md` exists:
+1. Read the ledger and the current phase plan
+2. Verify every **Must Have** feature has a non-empty entry in the **Phase** column
+3. If any Must Have feature has no phase assignment → **STOP**: "Must Have feature `[ID: name]` has no phase assignment. Assign it via `/plan-phase` before scaffolding."
+
+---
+
 ## 1. Read architecture pattern
 
 Read `docs/plans/*-architecture-design.md` and determine:
@@ -154,4 +163,16 @@ Run the dev server command from `.agent/instructions/commands.md`:
 
 **Pass criteria**: Dev server starts, serves a response, and shuts down cleanly.
 
+## 9.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 > Present result to user: "✅ Scaffold complete. Dev server starts cleanly. Proceeding to CI/CD setup." or "❌ Dev server failed: [error]. Fix required before continuing."
+
+> ❌ **NEXT STEP RESTRICTION**: The next workflow is `/setup-workspace-cicd`. Do NOT proceed to `/implement-slice` or any implementation work until all four setup-workspace shards (scaffold → cicd → hosting → data) have completed. Skipping setup shards will cause infrastructure failures during implementation.

package/template/.agent/workflows/setup-workspace.md

@@ -22,8 +22,9 @@ Operational setup of the project workspace, CI/CD pipeline, hosting platform, an
 
 ## 0. Pre-flight
 
-1. Read the approved phase plan from `docs/plans/phases/phase-N.md`
-2. Read `docs/plans/*-architecture-design.md` for the architecture pattern
+1. Read `.agent/skills/session-continuity/SKILL.md` and follow its session-open protocol. Check `.agent/progress/sessions/` for any previous session working on `setup-workspace`. If found → read the session close log to determine which shards completed and resume from the next incomplete shard.
+2. Read the approved phase plan from `docs/plans/phases/phase-N.md`
+3. Read `docs/plans/*-architecture-design.md` for the architecture pattern
 
 **Architecture pattern detection** — determines iteration strategy:
 
@@ -67,6 +68,18 @@ After all 4 shards complete, run `/verify-infrastructure` with trigger `workspac
 
 ---
 
+## 2.5. Completion Gate (MANDATORY)
+
+1. Update `.agent/progress/` — mark workspace setup as complete
+2. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+3. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+4. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 3. Completion
 
 Use `notify_user` to present results:

package/template/.agent/workflows/validate-phase-readiness.md

@@ -136,8 +136,29 @@ If the `dependency-auditing` skill is installed (`.agent/skills/dependency-audit
 
 **Pass criteria**: Zero HIGH/CRITICAL vulnerabilities in production dependencies.
 
+## 8.7. Feature ledger reconciliation
+
+If `docs/plans/feature-ledger.md` exists:
+1. Read the ledger and the current phase plan
+2. For each **Must Have** feature assigned to this phase:
+   - Verify the `Implemented` column is marked complete
+   - Verify the slice(s) that implement it are marked complete in `.agent/progress/phases/phase-N.md`
+3. If any Must Have feature assigned to this phase is not implemented → **STOP**: "Phase N has unimplemented Must Have features: [list]. Complete them via `/implement-slice` before marking the phase as validated."
+
+## 8.8. Boundary stub audit
+
+Grep the codebase for `BOUNDARY:` comments:
+1. If **0 results** → ✅ Pass — no active boundary stubs
+2. If results found → for each:
+   - Verify a linked tracking issue exists and is open
+   - Verify a sentinel test exists
+   - Verify the boundary is for a spec that genuinely doesn't exist yet
+   - If any boundary stub lacks a tracking issue OR the referenced spec now exists → **STOP**: "Active boundary stub at `[file:line]` — the referenced spec now exists. Replace the stub with a full implementation before marking this phase as validated."
+
 ## 9. Document results
 
+**Pre-append verification**: Before appending, verify `docs/audits/phase-N-validation.md` exists and contains a `## Spec Coverage` section from Step 5.8 (quality shard). If the file does not exist or the section is missing → **STOP**: "Spec coverage sweep did not complete. Re-run `/validate-phase-quality` Step 5.8 before appending readiness results."
+
 **Note on report file**: `docs/audits/phase-N-validation.md` is written progressively. Step 5.8 creates the file and appends the `## Spec Coverage` section. Step 9 appends all remaining sections. Do not recreate or overwrite the file in Step 9 — append only.
 
 - Test results and coverage
@@ -154,6 +175,18 @@ If the `dependency-auditing` skill is installed (`.agent/skills/dependency-audit
 - Migration verification status
 - Pass/fail verdict
 
+## 9.5. Completion Gate (MANDATORY)
+
+1. Update `.agent/progress/` — mark phase as validated
+2. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+3. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+4. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 10. Present results and next steps
 
 Read .agent/skills/verification-before-completion/SKILL.md and follow its methodology.

package/template/.agent/workflows/verify-infrastructure.md

@@ -194,6 +194,17 @@ Read `docs/plans/*-architecture-design.md` section `## Observability Architectur
 
 ---
 
+## 6.7. Spec pipeline integrity check
+
+1. Read `.agent/progress/spec-pipeline.md` — verify all IA/BE/FE columns show ✅ for all shards included in the current phase
+2. If `docs/plans/feature-ledger.md` exists:
+   - Read the ledger and verify no **Must Have** features have gaps in IA/BE/FE coverage
+   - If any Must Have feature has an empty column → **STOP**: "Feature `[ID: name]` is missing `[column]` coverage. Complete the spec before proceeding to implementation."
+
+> This is a lightweight check — it does not re-audit specs, just confirms the tracking is intact.
+
+---
+
 ## 7-8. Finalize report
 
 Finalize the report using the **Final Report** template from `.agent/skills/prd-templates/references/infrastructure-report-template.md`. Update the Verdict field to `✅ PASS` or `❌ FAIL` and fill in Failures and Next Steps sections.
@@ -202,6 +213,17 @@ The final report must include all eight gate checks: 0 (placeholder audit), 1 (C
 
 ---
 
+## 8.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+3. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 9. Present results
 
 Use `notify_user` to present the verification verdict:

package/template/.agent/workflows/write-architecture-spec-deepen.md

@@ -43,6 +43,7 @@ Re-read the complete draft (interactions + contracts + data model + access contr
 - Access rules that don't cover all interaction types
 - Edge cases that imply missing error codes in contracts
 - Event schemas that carry fields not in the data model
+- **Event consumer cross-references**: For each event's listed consumers, verify the consumer shard exists in `docs/plans/ia/index.md` and its `## Interactions` or `## Event Schemas` section references this event. If a consumer shard doesn't reference the event → add a cross-reference link in both directions. If a consumer shard doesn't exist → **STOP**.
 
 Fix every inconsistency found. Present findings to the user.
 
@@ -85,6 +86,26 @@ Read .agent/skills/technical-writer/SKILL.md for the spec writing step.
 
 Replace the skeleton sections in `docs/plans/ia/[shard-name].md` with the full content from all passes. Ensure all cross-shard dependencies are bidirectional.
 
+### 9.5. Spec complexity gate
+
+Count the total lines in the written spec file.
+
+| Lines | Action |
+|-------|--------|
+| **≤ 400** | ✅ Pass — proceed to Step 10 |
+| **401–500** | ⚠️ Warning — present to user: "This IA spec is [N] lines. Downstream BE/FE specs will expand further. Consider splitting if sections are independently testable." Proceed after acknowledgment. |
+| **> 500** | 🛑 **Hard stop** — "This IA spec is [N] lines and will likely exceed agent context capacity during BE/FE spec writing. Split this shard into two IA specs via `/decompose-architecture-validate` before proceeding." Present the largest sections with their line counts as split candidates. |
+
+> **Why these thresholds**: A 500-line IA spec typically expands to 800-1200 lines per BE/FE spec (contracts, schemas, validation rules, component details). Agent context windows degrade at > 1000 lines of spec content plus the workflow instructions.
+
+### 9.1. Post-write verification
+
+Re-read `docs/plans/ia/[shard-name].md` and verify:
+1. All required sections contain non-empty content (not just headers or `<!-- TODO -->` markers)
+2. File size is > 0 bytes
+3. No `<!-- TODO -->` markers remain (outside of intentional `[N/A]` sections)
+4. If any check fails → the write was incomplete. Retry the write operation.
+
 ## 10. Update IA index
 
 Change the shard's status from 🔲 to ✅ in `docs/plans/ia/index.md`.
@@ -126,6 +147,18 @@ Read `.agent/skills/session-continuity/protocols/ambiguity-gates.md` and run the
 
 **Hard gate**: Do NOT propose `/write-be-spec` until `/audit-ambiguity ia` scores 0% ambiguity.
 
+## 13.5. Completion Gate (MANDATORY)
+
+1. Update `.agent/progress/spec-pipeline.md` — mark IA column for this shard as complete
+2. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+3. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+4. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 14. Request review and propose next steps
 
 > [!CAUTION]

package/template/.agent/workflows/write-architecture-spec-design.md

@@ -121,7 +121,7 @@ Define for each entity: tables/collections, fields, types, relationships, indexe
 
 Write `## Data Models` to shard file immediately after confirmation. Follow write verification protocol.
 
-> **Decision recording**: For non-trivial data model decisions (schema approach, denormalization trade-offs, index strategy), read `.agent/skills/session-continuity/protocols/06-decision-analysis.md` and follow the **Decision Effect Analysis Protocol**.
+> **Decision recording**: For non-trivial data model or access control decisions, follow `.agent/skills/session-continuity/protocols/06-decision-analysis.md`.
 
 ## 5. Design access control
 
@@ -137,8 +137,6 @@ Read .agent/skills/security-scanning-security-hardening/SKILL.md and apply its a
 
 Write `## Access Control` to shard file immediately after confirmation. Follow write verification protocol.
 
-> **Decision recording**: For access control architecture decisions (role hierarchy, ownership model, escalation paths), read `.agent/skills/session-continuity/protocols/06-decision-analysis.md` and follow the **Decision Effect Analysis Protocol**. Record to `memory/decisions.md`.
-
 ## 5.5. Accessibility specifications
 
 Read `{{SURFACES}}` to determine the project's target surfaces.
@@ -157,6 +155,10 @@ Write `## Accessibility` to shard file immediately after confirmation. Follow wr
 
 Write `## Event Schemas` to shard file immediately after confirmation (if applicable). Follow write verification protocol.
 
+### 6b. Event consumer cross-reference (if events defined)
+
+For each event's listed consumers: verify consumer shard exists in `docs/plans/ia/index.md` and references this event. If consumer shard missing → **STOP**. Full cross-ref verification runs in the deepen shard.
+
 ## 7. Document edge cases
 
 Read .agent/skills/resolve-ambiguity/SKILL.md and follow its methodology.
@@ -183,6 +185,10 @@ For each referenced deep dive:
 2. If still a skeleton → write exhaustive subsystem detail: algorithms/state machines, technology choices with rationale, phasing strategy, data schemas, failure modes, integration contracts, performance, and security.
 3. Write immediately — do not wait until Step 8. The parent shard's summary + link is sufficient; the deep dive file IS the content.
 
+## 7.7. Completion Gate (MANDATORY)
+
+Scan for memory-capture triggers (rule: `memory-capture`). Write patterns/decisions/blockers to `memory/`. If none → confirm "No new entries." **Not skippable** — complete before `notify_user`.
+
 ## 8. Present all sections and request approval
 
 **Section completeness gate**: Before requesting approval, verify the spec file at `docs/plans/ia/[shard-name].md` contains ALL of the following sections with non-empty content (not just headers or `<!-- TODO -->` markers):

package/template/.agent/workflows/write-be-spec-classify.md

@@ -75,6 +75,10 @@ Build a **Referenced Material Inventory**.
 ### 5c. Read deep dives
 List `docs/plans/ia/deep-dives/`. Read each referenced deep dive in full. Extract key decisions, architectural constraints, data schemas.
 
+**Deep dive completeness check**: For each deep dive file referenced by the IA shard:
+- If the file contains `<!-- TODO -->`, `[TBD]`, or sections with only headers and no content → **STOP**: "Deep dive `[filename]` is still a skeleton. Run `/write-architecture-spec-design` Step 7.5 to fill it before proceeding with BE spec writing."
+- If the file does not exist → **STOP**: "Deep dive `[filename]` is referenced but missing. Create it via `/write-architecture-spec-design` Step 7.5."
+
 ### 5d. Read testability section
 If the shard has testability/acceptance criteria → read for performance targets and test requirements.
 
@@ -82,6 +86,16 @@ If the shard has testability/acceptance criteria → read for performance target
 
 Read any completed cross-cutting specs at `docs/plans/be/00-*.md`.
 
+## 6.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 7. Present classification and request approval
 
 Use `notify_user` presenting: classification, expected spec count, Referenced Material Inventory, split boundaries (if applicable).
@@ -91,3 +105,5 @@ Use `notify_user` presenting: classification, expected spec count, Referenced Ma
 After approval: read `.agent/skills/prd-templates/references/be-spec-template.md` → create spec file stub at `docs/plans/be/[NN-feature-name].md`.
 
 For structural reference (0 BE specs): confirm no write shard needed, propose next IA shard.
+
+> ❌ **NEXT STEP RESTRICTION**: Do NOT begin writing the BE spec (`/write-be-spec-write`) until the user has explicitly approved the classification and material inventory. If deep dive completeness issues were flagged in Step 5c, they must be resolved before proceeding.

package/template/.agent/workflows/write-be-spec-write.md

@@ -47,8 +47,22 @@ Read .agent/skills/testing-strategist/SKILL.md and follow its methodology.
 
 **Naming convention**: Use the same number prefix as the IA shard that sources it, followed by a kebab-case feature name. For multi-domain splits from the same shard, append a letter suffix (e.g., `09a-chat-api.md`, `09b-agent-flow-api.md`). For cross-cutting specs, use the `00-` prefix (e.g., `00-api-conventions.md`).
 
+**Split group tracking**: If this spec results from a split shard (letter suffix in filename), populate the `## Split Group` section in the spec with the split origin shard, companion spec filenames, and shared entity names. This is mandatory for split specs — it enables downstream implementation to discover sibling context.
+
 Read `.agent/skills/prd-templates/references/be-spec-template.md` for the document structure and quality gates checklist. Follow the conventions template from `be/index.md`.
 
+Write decision to disk. Continue below.
+
+### 7.5. Spec complexity gate
+
+Count the total lines in the written BE spec file.
+
+| Lines | Action |
+|-------|--------|
+| **≤ 600** | ✅ Pass |
+| **601–800** | ⚠️ Warning — "This BE spec is [N] lines. Consider splitting if endpoint groups are independently testable." Proceed after acknowledgment. |
+| **> 800** | 🛑 **Hard stop** — "This BE spec is [N] lines and will degrade implementation quality. Split the parent IA shard or separate endpoint groups into distinct BE specs." Present the largest sections with line counts. |
+
 ## 8. Update the BE index
 
 Add or update the spec entry in `docs/plans/be/index.md`. For multi-domain splits, add one row per BE spec with the shared IA source.
@@ -134,6 +148,22 @@ If this BE spec introduces a technology not already in the project's tech stack:
    - **2nd failure** → **STOP**: tell the user which dependency failed and ask: "Install manually, skip, or abort?"
 4. Confirm the matching skill is installed before proceeding.
 
+## 13.5. Update feature tracking ledger
+
+If `docs/plans/feature-ledger.md` exists, read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow **Step 3 — BE Coverage**. Match the endpoints in this spec to Feature IDs and populate the BE Spec and BE Status columns.
+
+## 13.7. Completion Gate (MANDATORY)
+
+1. Update `.agent/progress/spec-pipeline.md` — mark BE column for this shard as complete
+2. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+3. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+4. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 14. Request review and propose next steps
 
 > [!CAUTION]

package/template/.agent/workflows/write-fe-spec-classify.md

@@ -138,6 +138,16 @@ any that contain FE-relevant decisions.
 
 Read any completed cross-cutting FE specs — feature specs must follow their patterns. List the files matching `docs/plans/fe/00-*.md` (cross-cutting specs).
 
+## 5.5. Completion Gate (MANDATORY)
+
+1. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+2. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 6. Present classification and request approval
 
 Call `notify_user` presenting:

package/template/.agent/workflows/write-fe-spec-write.md

@@ -36,8 +36,22 @@ Read .agent/skills/testing-strategist/SKILL.md and follow its methodology.
 
 **Naming convention**: Numbered prefix matching feature position + kebab-case name (e.g., `01-auth-ui.md`). Cross-cutting: `00-` prefix.
 
+**Split group tracking**: If this spec results from a split shard (letter suffix in filename), populate the `## Split Group` section in the spec with the split origin shard, companion spec filenames, and shared entity names. This is mandatory for split specs — it enables downstream implementation to discover sibling context.
+
 Read `.agent/skills/prd-templates/references/fe-spec-template.md` for the document structure and quality gates checklist. Follow the conventions from `fe/index.md`.
 
+Write decision to disk. Continue below.
+
+### 6.5. Spec complexity gate
+
+Count the total lines in the written FE spec file.
+
+| Lines | Action |
+|-------|--------|
+| **≤ 600** | ✅ Pass |
+| **601–800** | ⚠️ Warning — "This FE spec is [N] lines. Consider splitting if component groups are independently testable." Proceed after acknowledgment. |
+| **> 800** | 🛑 **Hard stop** — "This FE spec is [N] lines and will degrade implementation quality. Split into separate FE specs per component group or page." Present the largest sections with line counts. |
+
 ## 7. Update the FE index
 
 Change the spec's status from 🔲 to ✅ in `docs/plans/fe/index.md`.
@@ -128,6 +142,22 @@ If this FE spec introduces a new technology:
 2. Read `.agent/workflows/bootstrap-agents.md` and invoke `/bootstrap-agents PIPELINE_STAGE=write-fe-spec` + the new dependency key
 3. **HARD GATE**: Follow the bootstrap verification protocol (`.agent/skills/prd-templates/references/bootstrap-verification-protocol.md`). Confirm the matching skill is installed before proceeding.
 
+## 12.5. Update feature tracking ledger
+
+If `docs/plans/feature-ledger.md` exists, read `.agent/skills/prd-templates/references/feature-ledger-protocol.md` and follow **Step 4 — FE Coverage**. Match the components in this spec to Feature IDs and populate the FE Spec and FE Status columns.
+
+## 12.7. Completion Gate (MANDATORY)
+
+1. Update `.agent/progress/spec-pipeline.md` — mark FE column for this shard as complete
+2. Scan this conversation for memory-capture triggers (see rule: `memory-capture`):
+   - Patterns observed → write to `memory/patterns.md`
+   - Non-trivial decisions made → write to `memory/decisions.md`
+   - Blockers hit → write to `memory/blockers.md`
+3. If no triggers found → confirm: "No new patterns, decisions, or blockers to log"
+4. Read `.agent/skills/session-continuity/protocols/05-session-close.md` and write a session close log
+
+> **This step is not skippable.** Do not call `notify_user` until all items above are complete.
+
 ## 13. Request review and propose next steps
 
 Read .agent/skills/verification-before-completion/SKILL.md and follow its methodology.