Gap Analysis & Roadmap

Grace is a design-complete, implementation-partial orchestration layer. This page documents the current state inventory, categorizes 33 gaps across 7 dimensions, and presents 5 strategic implementation bundles that will bring Grace from design to production.

Current State

What’s Done

• ✅ 8 agent context files — operational system prompts deployed to OpenClaw instance
• ✅ GRACE Protocol — SPEC-01 through SPEC-05 fully specified (SPEC-05 implemented in STRATT Block I)
• ✅ Architecture documentation — 19 comprehensive design documents in architecture/
• ✅ Council governance — 7 councils, 35 agents, 4 gate authorities, 9 protected agents
• ✅ Namespace resolution — stratt:// (11 domains, 5 types) and choco:// (2 namespaces) defined
• ✅ STRATT reference implementation — 43 units across 4 domains, 2,488 tests, 95%+ coverage

What’s Not Done

• ❌ SPEC-05 traces — format designed, zero implementation
• ❌ Quality scoring — weights defined (0.40 conformance, 0.35 completeness, 0.25 efficiency), zero code
• ❌ DSPy export — format designed, zero implementation
• ❌ Veritas gate — automated quality enforcement specified, not coded
• ❌ Self-improvement cycle — MIPROv2 integration designed, zero integration
• ❌ R2 atomic publish — local filesystem only; cloud registry not operational
• ❌ MERIDIAN content — documentation site scaffolded, src/content/docs/ empty
• ❌ Grace SDK/API — pure markdown workspace, no programmatic interface
• ❌ GRACE spec published — internally documented but not yet at stratt.dev

Component Inventory

Component	Location	Purpose	Status
Agent context files (8)	agent/	OpenClaw system prompt context	✅ Active — deployed
Protocol specification	architecture/GRACE-PROTOCOL-OVERVIEW.md	SPEC-01–SPEC-05 mapping	✅ Design complete
Ecosystem documentation	grace-hq/docs/	Public-facing docs site (Astro + Starlight)	🔄 TASKSET 4 in progress
Trace protocol	automation/trace-protocol.md	SPEC-05 trace format	⚠️ Design only
Campaign engine	automation/campaign-engine.md	10 campaign trails, 3 audience registers	⚠️ Design only
Marketing KPI tracker	architecture/MARKETING-KPI-TRACKER.md	HB-01/02/03 milestones	⚠️ All “Not started”
Primer system	GRACE-PROTOCOL-OVERVIEW.md lines 174–267	10-phase domain extension	✅ Neuro domain complete
Gate rules	config/gate-rules.yaml	Council governance config	✅ Consumed by STRATT
Namespace config	config/namespaces.yaml	Cross-namespace resolution	✅ Consumed by STRATT

---

Gap Categories & Details

Category 1: FOUNDATION — Core Protocol Completeness

Severity: Critical | Effort: Medium–Large | Dependency: Blocking Categories 2–7

Gap	Issue	Current State	Target State	Effort
F-01	Execution trace implementation	Format in trace-protocol.md; run.ts writes flat JSON to .stratt/runs/ — not SPEC-05 format	YAML traces at automation/traces/ with chain-level + step-level fields per SPEC-05	M
F-02	Quality scorer implementation	Weights defined (0.40/0.35/0.25); no code	Pluggable scorer in @stratt/cli/lib/scorer.ts	M
F-03	R2 atomic publish pipeline	Local filesystem registry only; R2 writes not operational	Fingerprint commit + R2 write with rollback on failure	M
F-04	ChocoBridgeResolver HTTP deployment	HTTP client code in choco-resolver.ts; Choco HQ endpoint not deployed	Choco HQ service live with /resolve endpoint	S
F-05	STRATT TAD conformance gaps	78% conformance; missing CRDT layer, incomplete agent schema	95% conformance; all CRDT features, complete agent schema	XL

Category 2: DEVELOPER EXPERIENCE — Docs, SDK, Onboarding

Severity: High | Effort: Small–Large | Dependency: Blocks marketing, customer adoption

Gap	Issue	Current State	Target State	Effort
D-01	No SDK or API for Grace	Pure markdown workspace; no programmatic interface	Python SDK + REST API with 8 endpoints	L
D-02	MERIDIAN content empty	apps/meridian/src/content/docs/ has 0 files	≥5 content files: getting-started, CLI reference, tutorials, examples	M
D-03	No getting-started documentation	README exists but no quickstart, tutorial, or examples	10-page getting-started guide with walkthroughs	S
D-04	GRACE spec not published	Internally documented; M1-01 target status “Not started”	Spec v1.0 rendered at stratt.dev with canonical URI	M
D-05	No public CLI documentation	21 commands documented internally (TOOLS.md) but not published	Full CLI reference with examples and exit codes	S

Category 3: QUALITY & RELIABILITY — Testing, Evaluation, Regression

Severity: High | Effort: Medium–Large | Dependency: Blocks BUNDLE-C

Gap	Issue	Current State	Target State	Effort
Q-01	DSPy export pipeline	JSONL format specified; MIPROv2 loop designed; zero code	stratt export-dspy command with filtering: min-score, date-range, version-range	M
Q-02	Regression detection automation	Veritas gate auto-trigger (delta > 0.05) designed; not implemented	Automated regression detection with quality score comparison across versions	M
Q-03	Grace has zero tests	Pure markdown = no executable code = no tests	Test suite for agent routing, memory management, tool orchestration	S
Q-04	No E2E chain execution validation	stratt run exists but no automated E2E test harness	Automated E2E test harness validating full chain execution	M

Category 4: OBSERVABILITY — Tracing, Monitoring, Dashboards

Severity: Critical–High | Effort: Small–Large | Dependency: Feeds BUNDLE-C

Gap	Issue	Current State	Target State	Effort
O-01	No SPEC-05 trace storage	automation/traces/ directory does not exist; run.ts writes to .stratt/runs/	Persistent SPEC-05 trace storage with structured query API	M
O-02	No quality score dashboard	MERIDIAN analytics page scaffolded but no data source	Real-time dashboard: quality scores over time, per-unit and per-chain	M
O-03	No gate resolution audit log	Gate state machine defined; no persistent audit trail	Persistent audit log with timestamp, approver, reason, affected unit	S
O-04	Heartbeat checks are aspirational	HEARTBEAT.md defines checks; relies entirely on OpenClaw runtime	Standalone health check mechanism (not dependent on OpenClaw)	M

Category 5: SECURITY — Auth, RBAC, Data Classification

Severity: Medium–High | Effort: Medium–Large | Dependency: Blocks production deployment

Gap	Issue	Current State	Target State	Effort
S-01	Auth/RBAC is Phase 3 groundwork	Framework scaffolded in packages/cli/src/lib/auth.ts; not production-ready	Token-based auth with role enforcement for publish/gate/council commands	L
S-02	No MFA/2FA	Choco HQ GAP-001 identifies as enterprise adoption blocker	OAuth 2.0 + MFA support for CLI and API	M
S-03	Data classification enforcement not automated	PUBLIC/INTERNAL/CONFIDENTIAL tiers defined; no automated enforcement	Automated redaction/classification enforcement in trace export	M

Category 6: INNOVATION — Competitive Moats & Differentiation

Severity: Medium | Effort: Small–Large | Dependency: Differentiator value

Gap	Issue	Current State	Target State	Effort
I-01	DSPy MIPROv2 optimization loop	Unique differentiator; entirely unimplemented	Python/TypeScript bridge for automated prompt refinement	L
I-02	Veritas gate (automated quality enforcement)	Novel concept described in trace-protocol.md; not coded	Auto-trigger gate on regression (delta > 0.05) before promotion	M
I-03	VAPA category definition not published	Coined internally; no published positioning or external reference	Published positioning paper + external reference on VAPA category	S
I-04	Self-improvement cycle (trace → refine)	Evaluation cycle documented; no automation	Full cycle: trace collection → quality analysis → MIPROv2 refinement → re-publish	L

Category 7: PRIMER — Multi-Domain Extension System

Severity: High | Effort: Medium–XL | Dependency: Scales STRATT to all 11 domains

Gap	Issue	Current State	Target State	Effort
P-01	7 of 11 declared domains empty	neuro, finance, nutrition, legal, film, artist, core have 0 units and 0 councils	All 11 domains populated with 20–40 units each; 9–11 councils operational	XL
P-02	No councils for non-dev/ops/docs domains	Only Pathfinder, Hermes, Athena exist	8 additional councils (finance, law, neuro, nutrition, film, artist, core, shared)	L
P-03	Primer pipeline not formalized as STRATT chain	10-phase pipeline exists as markdown; not STRATT units	stratt://core/chain/domain-extension-pipeline@1.0.0 with 10 phases + 9 gates	L
P-04	Grace not primer-aware	Agent context files have no reference to primer, multi-domain routing, or domain extension	Updated IDENTITY, TOOLS, SOUL with primer-aware routing logic	M
P-05	6 duplicate file pairs + stale HTML export	Veritas_* matches Primer_*; HTML export is outdated	Deduplicate, remove HTML, consolidate to single canonical markdown per phase	S
P-06	Bio domain extension not converted to STRATT units	Complete worked example exists as markdown (20–40 prompts); not YAML unit files	All bio prompts converted to STRATT YAML units with fingerprints	M
P-07	Primer docs use XML prompt format, not STRATT YAML	Handbook phases are <phase> XML blocks, not YAML prompt_body	Translate all Primer handbook XML to STRATT unit YAML format	L
P-08	No domain profile → STRATT unit automated pipeline	Manual process (paste profile + handbook into Claude session)	CLI integration: stratt extend --profile <questionnaire> --domain <name>	L

---

Priority Scoring Matrix

Each gap is scored across five dimensions:

Dimension	Weight	Scale
Business Impact	30%	1–10 (10 = transformative)
Technical Risk	25%	1–10 (10 = dangerous if unaddressed)
Effort Required	−20%	1–10 (10 = massive; higher = worse)
Dependencies	15%	1–10 (10 = blocks everything)
User Demand	10%	1–10 (10 = critical need)

Formula: Score = (Impact × 0.3) + (Risk × 0.25) − (Effort × 0.2) + (Deps × 0.15) + (Demand × 0.1)

Top 15 Gaps (by score)

Rank	Gap	Score	Severity	Bundle
1	D-04: GRACE spec published	5.75	HIGH	B
2	F-02: Quality scorer	5.45	CRITICAL	A
3	F-01: Execution traces	5.35	CRITICAL	A
4	P-03: Primer pipeline as STRATT chain	5.25	HIGH	E
5	O-01: SPEC-05 trace storage	5.15	CRITICAL	A
6	P-04: Grace primer-aware	5.05	HIGH	E
7	D-02: MERIDIAN content	4.80	HIGH	B
8	Q-02: Regression detection	4.55	HIGH	C
9	P-06: Bio domain → STRATT units	4.45	HIGH	E
10	Q-01: DSPy export pipeline	4.35	HIGH	C
11	I-02: Veritas gate	4.25	HIGH	C
12	F-03: R2 atomic publish	4.05	HIGH	B
13	O-03: Gate audit log	3.95	MEDIUM	A
14	I-01: DSPy MIPROv2 loop	3.90	HIGH	C
15	S-01: Auth/RBAC production	3.40	MEDIUM	D

---

Five Implementation Bundles

BUNDLE-A: “Trace Foundation”

Focus: Core protocol completeness + observability Duration: 3–4 weeks Depends on: Nothing — start immediately Unblocks: BUNDLE-C (quality loop), BUNDLE-D (dashboards)

Gaps: F-01 (5.35), F-02 (5.45), O-01 (5.15), O-03 (3.95)

Deliverables:

1. Create automation/traces/ directory with .gitignore for trace storage
2. Extend stratt run to emit SPEC-05 traces (chain-level + step-level fields)
3. Implement quality scorer in @stratt/cli/lib/scorer.ts (conformance 0.40 + completeness 0.35 + efficiency 0.25)
4. Add gate resolution audit records to trace output

Success criteria:

• stratt run <chain> produces YAML trace at automation/traces/YYYY-MM-DD-{slug}.yaml
• Every execution includes quality_score, token_counts, step-level details
• Gates include resolution audit (approved_by, wait_duration_ms)

Risk: Medium (modifies hot path in run.ts)

---

BUNDLE-B: “Publish Pipeline”

Focus: Developer experience + foundation Duration: 3–5 weeks Depends on: Nothing — run in parallel with BUNDLE-A Unblocks: Customer adoption, marketing launch

Gaps: D-04 (5.75), F-03 (4.05), D-02 (4.80), D-05 (3.15)

Deliverables:

1. Draft GRACE spec v1.0 (consolidate SPEC-01–SPEC-05 into publishable format)
2. Implement R2 atomic publish in @stratt/cli/lib/r2.ts (fingerprint commit + rollback on failure)
3. Populate MERIDIAN docs: getting-started, CLI reference, spec summary, examples
4. Generate full CLI reference for all 21 commands

Success criteria:

• GRACE spec v1.0 rendered at stratt.dev
• stratt publish writes to R2 with atomic rollback
• MERIDIAN docs ≥5 content files
• CLI reference covers all commands with examples

Risk: Low-Medium (spec writing is effort-intensive but technically straightforward)

---

BUNDLE-C: “Quality Loop”

Focus: Quality & reliability + innovation Duration: 4–6 weeks Depends on: BUNDLE-A (traces must be complete) Unblocks: Self-improvement cycle, competitive differentiation

Gaps: Q-02 (4.55), Q-01 (4.35), I-02 (4.25), I-01 (3.90)

Deliverables:

1. Implement regression detection: compare quality scores across versions; delta > 0.05 flags regression
2. Implement stratt export-dspy command with filtering (min-score, date-range, version-range, chain, exclude-gates)
3. Implement Veritas gate auto-trigger: regression detection → automatic gate before promotion
4. Wire DSPy MIPROv2 optimization loop: JSONL → MIPROv2 optimiser → tuned prompts

Success criteria:

• stratt run --compare shows quality delta against previous version
• stratt export-dspy produces valid JSONL per trace-protocol.md format
• Regression (delta > 0.05) auto-triggers Veritas gate
• MIPROv2 loop ingests JSONL and produces candidate prompt revisions

Risk: High (requires Python/TypeScript bridge; MIPROv2 is external dependency)

---

BUNDLE-D: “Platform Hardening”

Focus: Security, observability, developer experience Duration: 4–6 weeks Depends on: BUNDLE-B (MERIDIAN), BUNDLE-A (traces for dashboard) Unblocks: Production deployment, enterprise adoption

Gaps: S-01 (3.40), O-02 (3.15), F-04 (2.35), D-03 (3.25)

Deliverables:

1. Harden Auth/RBAC: token-based auth with role enforcement for publish/gate/council commands
2. Deploy Choco HQ API endpoint: enable CHOCO_RESOLVE_MODE=http in production
3. Build quality score dashboard in MERIDIAN: visualise trace quality over time
4. Write 10-page getting-started guide with walkthroughs

Success criteria:

• Unauthenticated requests rejected
• choco:// URIs resolve via HTTP in production
• Dashboard shows quality trends
• Getting-started guide covers install → validate → publish workflow

Risk: Low-Medium (mostly integration work)

---

BUNDLE-E: “Primer Integration”

Focus: Primer system formalization + multi-domain scaling Duration: 6–8 weeks Depends on: BUNDLE-A (traces), BUNDLE-B (spec published) Unblocks: Horizontal domain scaling, 11-domain ecosystem

Gaps: P-03 (5.25), P-04 (5.05), P-06 (4.45), P-07 (3.75), P-01 (2.60), P-02 (3.10)

Deliverables:

1. Convert 10-phase Primer pipeline to STRATT chain: stratt://core/chain/domain-extension-pipeline@1.0.0
2. Update Grace agent context (IDENTITY, TOOLS, SOUL) with primer-aware routing
3. Convert bio domain markdown example to STRATT units (20–40 YAML files with fingerprints)
4. Translate Primer handbook XML to STRATT unit YAML format
5. Create councils for remaining 8 domains (finance, law, neuro, nutrition, film, artist, core, shared)
6. CLI integration: stratt extend --profile <questionnaire> --domain <name>

Success criteria:

• Primer pipeline is executable STRATT chain
• Bio domain fully populated with YAML units + fingerprints
• Grace routes domain-specific requests to appropriate council
• stratt extend command works end-to-end

Risk: High (large scope; requires domain expertise for each extension)

---

Execution Timeline

Now                 Week 4              Week 8              Week 12            Week 16
├── BUNDLE-A (3-4w)────┤
│   Traces + scoring
│
├────────── BUNDLE-B (3-5w) ──────────┤
│           Spec published, MERIDIAN
│
│           ├──────────── BUNDLE-C (4-6w) ────────────┤
│           │             Quality loop, DSPy
│           │
│           ├────────────────── BUNDLE-D (4-6w) ────────────────┤
│                         Platform hardening
│
│                         ├──────────────── BUNDLE-E (6-8w) ───────────────────────┤
│                                         Primer integration, domain scaling

Critical path: A → C | B → D | E (latter half)

---

Metrics & Success Definition

Project completion is defined as:

• ✅ All Tier A gaps (score ≥ 5.0) resolved
• ✅ GRACE spec v1.0 published at stratt.dev
• ✅ MERIDIAN documentation site live with ≥10 content files
• ✅ SPEC-05 traces in production (all chains traced automatically)
• ✅ Quality scorer and regression detection operational
• ✅ Primer system formalized as STRATT chain
• ✅ Neuro domain fully published as STRATT units
• ✅ Grace agent context updated for all 5 bundles
• ✅ Zero critical security gaps
• ✅ CLI fully documented (public reference at stratt.dev)

---

See Also

• Protocol Layers — The Five SPECS
• Primer — Domain Extension System
• Protocol Overview