← Back to Tracker
Repository Inventory
20+
Python library modules
Initial Plan Modules (from Presentation) vs GitHub Reality
1. Collect — Requirements Capture via Structured Templates
Customers fill Excel templates: territories, quotas, rules
85% Complete
Covered: Intake template structure documented (CONTEXT/08_INTAKE_TEMPLATE.md). Intake reader module built (spiff_setup/intake_reader.py). Krayden config workbook with Val's comments is the active source. Client patterns catalog (CONTEXT/06_CLIENT_PATTERNS.md). Intake template generator tool (tools/generate_intake_template.py).
Gap: No client-facing application/portal for self-service requirements input. Still Excel-based. "Consultant agent" for client-facing collection is conceptual only.
CONTEXT/08_INTAKE_TEMPLATE.md | CONTEXT/06_CLIENT_PATTERNS.md | spiff_setup/intake_reader.py | tools/generate_intake_template.py | REQUIREMENTS/Krayden_*.xlsx
2. Validate — Team Reviews and Validates Submitted Data
Review and validate customer-submitted data
90% Complete
Covered: Gap check skill (.claude/skills/spiff-gap-check/SKILL.md) audits requirements vs inventory automatically. Validator module (spiff_setup/validator.py). Session start checklist enforces gap audit before any work. Pre-bash hook blocks wrong org. Platform knowledge doc covers what API can/cannot do. Lessons learned doc captures anti-patterns.
Gap: No formal client sign-off workflow. Gap check is internal — no client-facing validation report.
.claude/skills/spiff-gap-check/SKILL.md | spiff_setup/validator.py | CONTEXT/01_PLATFORM_KNOWLEDGE.md | CONTEXT/05_LESSONS_LEARNED.md | scripts/hooks/verify_spiff_org.py
3. Configure — Import & Configure in Salesforce SPM
Agent-assisted Salesforce setup & configuration
80% Complete
Covered: Full Python library (20+ modules): auth, plans, fields, tables, scopes, filters, datasheets, layouts, payroll periods, quotas, users, teams. 8-phase setup procedure (CONTEXT/02_SETUP_PROCEDURE.md). Order of operations codified (0A Admin → 0B Users → 0C Teams → 0D Plans → 0E Quotas → 0F Data → 0G Designer). Browser automation module. Computer Use skill + Playwright skill. Designer configuration guide. API reference (CONTEXT/04_API_REFERENCE.md). GraphQL mutations. PE fully configured with 130 inventory entries. Krayden v3 plan doc with 40-step granular execution.
Gap: Import formatting issue (strings vs numbers). Sandbox-to-production gap unknown. Some formulas require SPIFF support tickets.
spiff_setup/*.py (20 modules) | CONTEXT/02_SETUP_PROCEDURE.md | CONTEXT/04_API_REFERENCE.md | CONTEXT/03_FORMULA_REFERENCE.md | SETUP/00-06_*.md | .claude/skills/spiff-computer-use/ | .claude/skills/spiff-playwright/
4. Verify — Check Outputs Match Excel Calculations
Automated output verification against source data
75% Complete
Covered: 32 sanity test scripts across 6 phases (scripts/sanity/phase_0 through phase_6). AI Test Agent with 8 checks: login, org binding, designer loads, teams, quotas, data sources, reporting, admin. Two-layer sanity: state assertion scripts + UI smoke via Computer Use. Pre-handoff validation command. Per-rep variance checker (phase_4_per_rep_variance.py). Hand-compute validator (phase_5_hand_compute.py).
Gap: No full-quarter commission comparison with live data. Regression testing defined but not fully populated. QA agent is conceptual (discussed in meeting, not built as standalone agent).
scripts/sanity/*.py (32 files) | AI_TEST_AGENT/ (8 checks + drivers) | AI_TEST_AGENT/regression_test.md | AI_TEST_AGENT/sanity_test.md | spiff_setup/validator.py
5. Document — Produce Project Documentation & Artifacts
Auto-generated specs, scoping docs & project plans
40% Complete
Covered: As-built inventory auto-generated (CSV + MD). Executive summaries generated (3 versions + Krayden-specific). Setup plan document (v3). Release notes generator (scripts/generate_release_notes.py). DOCX generators (scripts/make_krayden_plan_docx.py, generate_docx.py). PPTX generator. Configuration documentation generated during build (40-step log).
Gap: No client-facing technical documentation template. No solution design document generator. No role-based user guides (End User / Manager / Admin). No standardized deliverable package. Mohammad identified this as 15-20% complete in the meeting.
scripts/generate_release_notes.py | scripts/make_krayden_plan_docx.py | scripts/generate_docx.py | scripts/create_pptx.py | SETUP/_archive/*/EXECUTIVE_SUMMARY_*.md
6. Train — Deliver Role-Based Training to the Client
Role-based training materials & video templates
15% Complete
Covered: SPIFF AI Training Guide with build standards (REQUIREMENTS/SPIFF_AI_Training_Guide 1.md). Training added-value doc (SETUP/SPIFF_AI_TrainingGuide_AddedValue.docx). Demo video script (scripts/make_demo_video.py). Max Votek proposed automated video generation with avatars (discussed Apr 28).
Gap: This is the biggest gap. No role-based user guides (End User / Manager / Admin). No training video templates. No train-the-trainer materials. No in-app help content. No Salesforce-embedded learning paths. The presentation envisioned "role-based user guides" and "training materials & video templates" — neither exists as a reusable artifact.
REQUIREMENTS/SPIFF_AI_Training_Guide 1.md | SETUP/SPIFF_AI_TrainingGuide_AddedValue.docx | scripts/make_demo_video.py
7. Institutional Memory — Reusable Knowledge Base
Structured project templates, searchable knowledge base of past work
80% Complete
Covered: CONTEXT/ folder with 10 structured knowledge files. Platform knowledge, formula reference, API reference, lessons learned, client patterns, automation stack, org settings checklist (127 items), official docs inventory with Trailhead URLs. Quick-start procedure for new clients (10-step). 3 archived project snapshots (training Apr 17, build Apr 26, PE May 8/12). Prompt template for new sessions. /compact-spiff skill to regenerate knowledge from source.
Gap: Not yet searchable (no vector DB or search index). No cross-project analytics. Knowledge is file-based, not queryable. No formal template for post-project retrospective.
CONTEXT/*.md (10 files) | CONTEXT/prompt.md | CONTEXT/09_ORG_SETTINGS_REFERENCE.csv | SETUP/_archive/ (3 snapshots) | SAMPLES/_archive/ (historical guides)
8. Expand — AI-Driven Account Analysis for Upsell
Structured discovery to surface upsell & expansion opportunities
25% Complete
Covered: Org diagnostics / health check explored by Max Shevchenko (Apr 21). Pre-sales estimator built by Kevin (Apr 28, expanding). Schema discovery module (spiff_setup/schema_discovery.py). Inventory module can audit any org. Health check concept presented to Salesforce team. Two AEs requesting health checks.
Gap: No formal health check agent or report template. No structured discovery process. No upsell recommendation engine. Estimator is still in Excel (not converted to web app). No success story templates or industry-specific positioning. Website concept discussed but not built.
spiff_setup/schema_discovery.py | spiff_setup/inventory.py | scripts/probe_*.py
Critical Gaps Summary
| Priority | Gap | Impact | Planned Module | Effort |
|---|
| P1 |
Role-based user guides — End User / Manager / Admin guides not templated |
Clients asking for this NOW (Dmytro Terniak's project). Blocks "Document" and "Train" stages. |
Train + Document |
Medium |
| P1 |
Training video templates — No automated video generation pipeline |
Max Votek proposed approach with avatars. Prototype with Maksim Shcherbak "in a few hours." Not started. |
Train |
Medium |
| P2 |
Client-facing requirements portal — Still Excel-only intake |
Solovyov suggested Shiny app. Max Votek offered to convert. Kevin building estimator. No web form yet. |
Collect |
High |
| P2 |
Health check agent — Org assessment for existing clients |
2 AEs requesting. Opens managed services revenue. Concept only. |
Expand |
Medium |
| P2 |
Solution design document generator — No automated spec creation |
Executive summaries exist but no full solution design. Client sign-off document missing. |
Document |
Low |
| P3 |
Pre-sales estimator as web app — Still Excel |
Kevin target: end of this week. Solovyov can convert. Mohammad presenting to Salesforce. |
Expand |
Low |
| P3 |
QA agent — Standalone quality assurance agent |
Sanity scripts exist. Test agent exists. But no autonomous QA agent that validates independently. |
Verify |
Medium |
| P3 |
Searchable knowledge base — Files only, no search/query |
10 MD files + prompt template. Works for Claude context. Not searchable by humans or across projects. |
Institutional Memory |
Medium |
What's Surprisingly Strong (Exceeds Initial Plan)
Automation depth. The initial plan said "AI-assisted configuration." The repo has a full Python library (20 modules), CLI entry point, 3 Flask web UIs, 32 sanity scripts, Computer Use integration, Playwright integration, Docker setup, pre-bash hooks, and 4 Claude skills. This is far beyond "assisted" — it's a configuration engine.
Safety architecture. Two-layer sanity (script + UI smoke), pre-bash hook blocking wrong org, rollback at every step, org cleaner agent with dry-run + approval gate. This wasn't in the initial plan at all.
Knowledge compaction. The CONTEXT/ folder with 10 structured files replacing hundreds of pages of training material is exactly what the "institutional memory layer" envisioned — and it's already working with a 10-step quick-start for new clients.
Agent specialization. Two working agents (test + cleanup) with web UIs. The initial plan mentioned "long-horizon coding agent" as singular. The repo has moved to specialized multi-agent architecture.
Analysis: github.com/maxshev2000github/SPIFF_SETUP_V2 vs AI-Driven Salesforce Delivery Factory presentation