Skip to content

RF — TFW-18 / Phase B: Knowledge Quality

Дата: 2026-04-03 Автор: Executor (AI) Статус: 🟢 RF — Выполнено Parent HL: HL Phase B TS: TS Phase B

1. Что сделано

Новые файлы

Файл Описание
ONB__PhaseB__knowledge_quality.md Executor onboarding report (0 blocking questions)

Изменённые файлы

Файл Изменения
.tfw/templates/RF.md §6 FC: reframed «project knowledge → strategic knowledge», swapped anti-patterns for good/bad examples
.tfw/templates/REVIEW.md §5 FC: same reframe as RF.md
.tfw/templates/RES.md §FC: same reframe + kept «THIS project» filter, added tfw-docs redirect for impl details
.tfw/workflows/research.md §Closure L110: one-line reframe to strategic knowledge
.tfw/workflows/handoff.md L81-87: leads with «stakeholder priorities, domain patterns» instead of «environment, constraints»
.tfw/conventions.md §10.1: expanded category examples (domain→revenue patterns/client segments, stakeholder→priorities/pain points, constraint→contractual obligations, etc.)
.tfw/workflows/knowledge.md Phase 2: «strategic knowledge», mentions «decisions, emotions, data», redirects tech details to tfw-docs
.tfw/README.md §v3: added Knowledge consolidation bullet; workflows table: added knowledge + config rows
.agent/workflows/tfw-research.md Synced from canonical (cp)
.agent/workflows/tfw-handoff.md Synced from canonical (cp)
.agent/workflows/tfw-knowledge.md Synced from canonical (cp)

2. Ключевые решения

  1. Exact TS text used: All replacement text taken verbatim from TS Steps 1-8. No deviations.
  2. Adapter sync via cp: Same method as Phase A (byte-identical copies). 3 adapters synced.

3. Acceptance Criteria

  • [x] RF.md FC prompt says «strategic knowledge» not «next agent's behavior»
  • [x] REVIEW.md FC prompt matches RF.md
  • [x] RES.md FC prompt same reframe + keeps «THIS project» filter
  • [x] research.md §Closure FC bullet reframed
  • [x] handoff.md FC guidance leads with «stakeholder priorities, domain patterns»
  • [x] conventions.md §10.1 domain examples include: revenue patterns, client segments, market metrics
  • [x] conventions.md §10.1 stakeholder examples include: priorities, pain points, quotes
  • [x] knowledge.md Phase 2 says «strategic knowledge» and mentions «decisions, emotions, data»
  • [x] .tfw/README.md §v3 lists Knowledge consolidation
  • [x] .tfw/README.md workflows table includes knowledge and config
  • [x] All 3 adapters synced (diff = 0)
  • [x] No file exceeds P10 word limit after changes (research.md: 1165w, knowledge.md: 718w)
  • [x] Anti-pattern examples include «implementation details» as NOT-FC

4. Верификация

  • Lint (echo "configure your lint command"): N/A (markdown files only)
  • Tests (echo "configure your test command"): N/A (markdown files only)
  • Verify: All grep/diff checks passed:
  • strategic knowledge found in RF.md, REVIEW.md, RES.md, research.md, knowledge.md
  • next agent's behavior count = 0 in RF.md, REVIEW.md
  • revenue patterns and priorities, pain points found in conventions.md
  • Knowledge consolidation found in README.md
  • All 3 adapters: diff = 0
  • Word counts: research.md 1165w (≤1200), knowledge.md 718w (≤1200)

5. Observations (out-of-scope, not modified)

# File Line(s) Type Description
1 .tfw/README.md 77 todo Workflows table in §Project Structure tree still lists old set (plan, research, handoff, review, resume, docs, release, update) — doesn't include knowledge/config. Only the table in §Canonical Workflows was updated per TS scope
2 .tfw/workflows/review.md — style review.md also has a FC mindset reminder (added in Phase A) but was not in scope for Phase B reframe — still uses old «project knowledge» framing

Types: dead-code, naming, todo, duplication, perf, security, style, missing-test, ux

6. Fact Candidates

Before writing Fact Candidates, review the conversation history. The human's messages are the primary source of strategic knowledge — domain insights, stakeholder priorities, business context, and constraints that shape decisions.

Good: "18% clients = 80% revenue (Pareto)", "stakeholder: find problem clients first" NOT fact candidates: "project uses git", "code is in Python", implementation details (→ §5 Observations → tfw-docs) These are NOT verified facts. They become facts after /tfw-knowledge consolidation.

# Category Candidate Source Confidence
1 process Phase B "knowledge quality" reframe was a pure text-edit task (0 new files, 8 modifications, ~120 words) — this size is optimal for single-session executor completion Execution timing Medium
2 convention The .tfw/README.md has two places listing workflows: a tree in §Project Structure (L77) and a table in §Canonical Workflows (L161-172). Both need updating when workflows change — easy to miss the tree Observation #1 during execution High

Categories (open list): environment, process, stakeholder, constraint, convention, domain, context, risk

RF — TFW-18 / Phase B: Knowledge Quality | 2026-04-03