trace-first-starter

Trace-First Workflow (TFW) — Canonical Starter

“The thought process and the instructions are more valuable than the immediate result.”

Most work with AI happens in dialogue — and then the chat ends. Tomorrow you re-explain the project from scratch. Context evaporates across sessions, models, and teams. TFW solves this: a single ritual that captures intent, structures decisions, and makes any project resumable by any agent or human.

For the full philosophy, thesis, and design rationale → .tfw/README.md


Who This Is For

TFW is domain-agnostic. It works for:

If your work involves AI-assisted iteration and you need continuity across sessions — TFW is for you.


Quick Start

New project

  1. Fork or clone this repository
  2. Follow .tfw/init.md to configure your project
  3. Choose your tool adapter (Claude Code, Cursor, Antigravity, or plain chat)
  4. Start with .tfw/workflows/plan.md to create your first task

Mid-conversation

Already deep in a chat and realize it’s time to get structured?

  1. Paste this repo link: https://github.com/saubakirov/trace-first-starter
  2. The agent reads README → AGENTS, harvests your existing context, and bootstraps TFW files
  3. Review the drafts, pick a mode, continue — now with traces

What’s Inside

Root Files (your project)

File Purpose
README.md Project guide + Task Board
AGENTS.md AI agent role and behavior
KNOWLEDGE.md Architecture, decisions, legacy index
TECH_DEBT.md Tech debt registry
RELEASE.md Release strategy and context (optional)

.tfw/ (TFW core — tool-agnostic)

Path Contents
.tfw/README.md Philosophy, thesis, lifecycle, anti-patterns, evolution
.tfw/conventions.md Formal rules, statuses, naming, scope budgets
.tfw/glossary.md Terminology
.tfw/templates/ Canonical templates (HL, TS, RF, ONB, REVIEW)
.tfw/workflows/ Process workflows (plan, handoff, resume, release, update)
.tfw/adapters/ Tool adapter templates
.tfw/init.md Setup instructions
.tfw/PROJECT_CONFIG.yaml Project parameters
.tfw/VERSION Current framework version (semver)
.tfw/CHANGELOG.md Version history

Tool Adapters

TFW works with any development tool. Templates in .tfw/adapters/:

Tool Adapter Project entry point
Claude Code .tfw/adapters/claude-code/ CLAUDE.md (project root)
Cursor .tfw/adapters/cursor/ .cursor/rules/tfw.mdc
Antigravity .tfw/adapters/antigravity/ .agent/rules/tfw.md
Plain chat Read .tfw/README.md directly

Setup details in .tfw/init.md.


Key Concepts



Task Board

ID Task Status HL TS ONB RF REV  
TFW-1 Formalize success criteria ✅ DONE  
TFW-2 Upgrade to TFW v3 ✅ DONE  
TFW-3 Root README public-readiness 🟢 RF    
TFW-4 Framework cleanup 🟡 TS        
TFW-5 KNOWLEDGE.md + tfw-docs workflow ✅ DONE  
TFW-6 Versioning, changelog, tfw-update workflow ✅ DONE A B C A B C A B C  
TFW-7 Resolve all open tech debt ✅ DONE  
TFW-8 Reviewer role + /tfw-review workflow ✅ DONE A B A B  
TFW-9 Update source mechanism for tfw-update ✅ DONE  
TFW-10 Replace stale “TFW v3” labels with semver ✅ DONE  
TFW-11 RESEARCH stage in pipeline ✅ DONE A B C A B C  
TFW-12 Centralize config params in PROJECT_CONFIG ✅ DONE  
TFW-13 tfw-init workflow (replace init.md) ✅ DONE A B A B A B A B  
TFW-14 Research interaction model (briefing + handoff) ✅ DONE  
TFW-15 Pipeline rename: separate statuses from documents (HL_DRAFT → RES → TS_DRAFT) ✅ DONE  
TFW-16 tfw-doctor: analyze user behavior, chats, KI, tasks — find missed workflows, knowledge gaps, hidden flows ⬜ TODO            
TFW-17 Research depth + coordinator quality (skip-bias, external tools, rush-bias) ✅ DONE  
TFW-18 Knowledge consolidation: fact candidates, dream-like docs, mandatory gate ✅ DONE  
TFW-19 tfw-config: propagate PROJECT_CONFIG.yaml changes to workflows/adapters automatically ✅ DONE  
TFW-20 tfw-user-tune: personal preferences pipeline (.user_preferences.md lifecycle, gitignored, user-specific) ⬜ TODO            
TFW-21 Compress research.md: 2397→1145 words (-52%), deduplicate, remove inline templates ✅ DONE  
TFW-22 Coordinator & Research enrichment: result visualization in HL, research justification, structured thinking algorithms ✅ DONE  
TFW-23 Templates English standardization: eliminate mixed RU/EN, pure English templates + content_language config ✅ DONE  
TFW-24 RES State Machine: Researcher role, subfolder state machine, resume protocol, HL Vision/Impact, Working Backwards ✅ DONE A B A B A B A B
TFW-25 Values & Principles consolidation: enrich README Values, prune KNOWLEDGE.md, clean knowledge/ facts ✅ DONE
Statuses: ⬜ TODO → 📝 HL_DRAFT → 🔬 RES → 🟡 TS_DRAFT → 🟠 ONB → 🟢 RF → 🔍 REV → ✅ DONE ❌ BLOCKED