Skip to content

title: "Extract — "What do we NOT see?"" source: "tasks/TFW-25__values_consolidation/research/extract.md"


Extract — "What do we NOT see?"

Parent: HL-TFW-25 Goal: Exact audit — which P# items move, which knowledge/ facts are prunable, which README values stay/change.

Findings

E1: P1-P14 classification with action

Cross-referencing every P# against the Values/Principles/Rules taxonomy and checking if the content is already expressed elsewhere:

P# Content (short) Tier Already in... Action
P1 Traces over code Value README §Values ("Traces Over Code") STAY in §0. Already in README — no change needed
P2 Index, don't duplicate Value README §Values ("Single Source of Truth") STAY in §0. Subsumed by SSOT in README
P3 Philosophy stays rich — narrative > DRY Value Implicit in README narrative format STAY in §0. Unique nuance — worth keeping
P4 Glossary = dictionary, conventions = rules Rule Self-evident from conventions.md §1 + glossary format REMOVE from §0. Obvious from file structure
P5 Meta-project awareness Principle Unique — not stated elsewhere KEEP in §0 (compress to 1-liner). Relevant for meta-project only
P6 Lightweight docs Rule Implicit in tfw-docs workflow (5-item checklist) REMOVE from §0. Workflow IS the implementation
P7 Self-review ≠ review — separate roles Value README §Roles (4 roles). conventions §15 (Role Lock) STAY in §0. Core belief about quality
P8 RESEARCH ≠ passive checklist Value/Principle research/base.md Mindset + Rules COMPRESS in §0. Long description → 1-liner. Enforcement lives in workflow
P9 Coordinator Mindset: quality > speed Value README §Values ("Completeness Over Speed" — partial). plan.md Mindset section MERGE into README "Completeness Over Speed" or keep as §0 1-liner
P10 Token density ≤1200 words Rule constraint/F2. base.md limits table. PROJECT_CONFIG MOVE to conventions or REMOVE from §0. Pure engineering constraint
P11 Enforcement values MUST be inline Rule D24 (Config Sync Registry). convention/F5 MOVE to conventions or REMOVE from §0. Engineering pattern
P12 DNA / Library split Rule D25. Convention/F5 (ref-inside-step) MOVE to conventions or REMOVE from §0
P13 Progressive Disclosure Principle D25 (modular architecture). Implicit in research/{base,focused,deep}.md REMOVE from §0. The architecture IS the principle
P14 Filesystem = state machine Value philosophy/F4. conventions §4. research/ subfolder PROMOTE to README as "Structural Enforcement". Remove from §0 after promotion

Summary: - STAY in §0 (compact): P1, P2, P3, P5, P7, P8, P9 (7 items) - REMOVE from §0: P4, P6, P10, P11, P12, P13 (6 items) - PROMOTE to README: P14 (1 item → "Structural Enforcement")

E2: knowledge/ facts self-evidence audit

Checking each fact: "Is this now obvious from the code it describes?"

convention.md (12 facts)

# Fact (short) Self-evident from... Verdict
F1 Adapter = byte-copy Adapter files themselves ⚠️ Keep — not obvious without trying
F2 CRLF inconsistency No rule exists → still relevant Keep
F3 TS descriptions can drift General principle, not code-specific Keep
F4 Checkpoint fields in templates templates/research/*.md PRUNE — templates show this
F5 Ref-inside-step pattern plan.md, base.md demonstrate it PRUNE — code IS the pattern
F6 Config Sync Registry scope config.md workflow explains it PRUNE — workflow shows this
F7 Adapter drift → copy-on-modify Related to F1 Keep (complements F1)
F8 §3.1 domain-agnostic templates/HL.md shows it PRUNE — template IS the rule
F9 Filesystem state machine conventions §4, research/ subfolder PRUNE — P14 + architecture show this
F10 RES = synthesis templates/RES.md structure PRUNE — template IS the proof
F11 HL Working Backwards templates/HL.md §1 Borderline — keep (template doesn't explain WHY)
F12 Stage templates structure templates/research/* PRUNE — templates exist

Prunable convention facts: F4, F5, F6, F8, F9, F10, F12 = 7 facts Keep: F1, F2, F3, F7, F11 = 5 facts

process.md (10 facts)

# Fact (short) Self-evident from... Verdict
F1 TFW = teamwork README §Roles Borderline — philosophical, keep
F2 tfw-knowledge ≠ tfw-docs Two separate workflows exist PRUNE — file existence = documentation
F3 Scan conversation for FCs templates/RES.md, RF.md have the instruction PRUNE — template says it
F4 Best FCs from user emotions Strategic insight — NOT in any code Keep
F5 Naming creates behavior D28, philosophy — user confirmed = Value Keep (might move to philosophy/)
F6 Agents follow steps + gates Process insight — not directly in code Keep
F7 Write to file first, then chat Process insight — not enforced in code Keep
F8 Crash recovery / gate skipping D31, process/F10 covers the fix PRUNE — F10 (resume protocol) covers the problem + solution
F9 4 roles conventions §15, glossary §Roles PRUNE — obvious from code
F10 Resume Protocol base.md Step 0 PRUNE — code IS the protocol

Prunable process facts: F2, F3, F8, F9, F10 = 5 facts Keep: F1, F4, F5, F6, F7 = 5 facts

philosophy.md (4 facts)

# Fact (short) Self-evident? Verdict
F1 Zettelkasten inspiration Strategic origin story Keep
F2 tfw-docs vs tfw-knowledge = different systems Overlaps with process/F2 Keep (this is the authoritative version)
F3 User wants critical opponent Strategic — user preference Keep
F4 Structural enforcement > format enforcement P14 + README value Keep but compress (will be in README)

Prunable philosophy facts: 0 (user confirmed: philosophy = keep)

constraint.md (3 facts)

# Fact (short) Self-evident? Verdict
F1 User prefs not in shared files Not in any code Keep
F2 Workflow degrades >1200 words P10 + PROJECT_CONFIG Borderline — keep (empirical finding)
F3 Agents generate filler FCs Process insight Keep

Prunable constraint facts: 0

E3: README Values — proposed final list

Current (5 values): Candor, Completeness, Determinism/Safety, Portability, SSOT

Proposed (7 values):

# Value Status Source
1 Traces Over Code Exists (implied in thesis, not in §Values) §Thesis
2 Candor Over Flattery Exists — enrich with P9 coordinator mindset §Values + P9
3 Completeness Over Speed Exists §Values
4 Structural Enforcement NEW — from P14, philosophy/F4 P14 + TFW-24
5 Naming Creates Behavior NEW — from D28, process/F5, user belief D28 + TFW-22
6 Single Source of Truth Exists §Values
7 Portability Exists §Values

Dropped: "Determinism and Safety" — fully covered by conventions §12. Not a belief, it's a rule set (don't fabricate, don't guess, use env vars). HL flagged this as merge candidate.

Count: 7 — within the 4-8 range validated by external research.

E4: KNOWLEDGE.md §3 Legacy pruning candidates

Current: 35 items (lines 115-148). Items where Status = Removed/Replaced AND the replacement is fully implemented:

Checking which items are now purely historical with no active reference value:

  • Lines 115-121 (STEPS.md, TASK.md, Summary Discipline, AI_ENTRY_POINT, SUCCESS_CRITERIA, 00_meta/, Review in handoff) — all from TFW-2/4/8, pre-v3. No one references these anymore. PRUNE all 7.
  • Lines 123-125 (inline scope budgets, version strings, hardcoded template lists) — TFW-12 era. Superseded mechanics. PRUNE all 3.
  • Lines 127-128 (Complexity Check, status emoji rename) — TFW-14/15. Done. PRUNE 2.
  • Lines 132-135 (Example Flow, good/bad research, inline checkpoint, duplicate anti-patterns) — TFW-21. All removed. PRUNE 4.
  • Line 136 (Pattern B superseded) — D17D24. Already in D24. PRUNE 1.
  • Line 137 (Naming rules removed from plan.md) — TD-48 cleanup. PRUNE 1.

Total prunable Legacy items: 18 out of 35 (~51%) Keep: 17 recent items (TFW-22 through TFW-24)

E5: KNOWLEDGE.md §4 Tech Stack

4 lines of content: Markdown+YAML, Git/GitHub, Claude/Cursor/Antigravity, MIT License. All trivially obvious from: repo files, .tfw/adapters/, LICENSE file, README.

Verdict: REMOVE entire section. (HL already proposed this.)

Checkpoint

Found Remaining
P1-P14 classified: 7 stay, 6 remove, 1 promote None
12 knowledge/ facts prunable (7 convention + 5 process) Need to verify no cross-refs break
README Values: 7 items (drop Determinism/Safety, add 2 new) None
Legacy: 18 items prunable None
§4 Tech Stack: remove None

Sufficiency: - [x] External source used? (taxonomy from Gather validates classification) - [x] Briefing gap closed? (exact lists produced for all pruning targets)

Stage complete: YES → User decision: ___