RF — {PREFIX}-{N} / Phase {X}: {Title}¶
Date: YYYY-MM-DD Author: {author} Status: 🟢 RF — Complete Parent HL: HL-{PREFIX}-{N} TS: TS Phase {X}
1. What Was Done¶
New Files¶
| File | Description |
|---|---|
path/to/file |
{description} |
Modified Files¶
| File | Changes |
|---|---|
path/to/file |
{description} |
2. Key Decisions¶
- {Decision and rationale}
- {Decision and rationale}
3. Acceptance Criteria¶
- [x] {Criterion from TS}
- [x] {Criterion from TS}
4. Verification¶
- Lint (
{config.build.lint}): {result} - Tests (
{config.build.test}): {result} - Verify (
{config.build.verify}): {result}
5. Observations (out-of-scope, not modified)¶
| # | File | Line(s) | Type | Description |
|---|---|---|---|---|
| 1 | path/to/file |
{lines} | {type} | {description} |
Types:
dead-code,naming,todo,duplication,perf,security,style,missing-test,uxQuality bar: report only issues that would bite the next developer. Don't generate observations just because the section exists. If nothing found:
No observations.
6. Fact Candidates¶
Cognitive mode: Pure reporting — record factual observations without interpretation or synthesis.
Scope: Agent-observed project patterns discovered during execution. Good: "18% clients = 80% revenue (Pareto)", "stakeholder: find problem clients first" NOT fact candidates: "project uses git", implementation details (→ §5 Observations → tfw-docs), or agent-generated analysis (→ §7 Strategic Insights).
Human-Only Test: would this fact be unknown without the human saying it? If an agent can discover it by reading code or running commands — it's not a fact candidate. These are NOT verified facts. They become facts after
/tfw-knowledgeconsolidation.Before writing: review the conversation history. The human's messages are the primary source.
| # | Category | Candidate | Source | Confidence |
|---|---|---|---|---|
| 1 | {category} | {what you learned} | {where from} | High/Medium/Low |
Source format: Use reference patterns (e.g.,
[RF TFW-18](../../tasks/TFW-18__knowledge_consolidation/RF__PhaseB__knowledge_quality.md),D24). See compilable_contract.md §2.Categories (open list): see conventions.md §10.1 for full list with scope descriptions.
7. Strategic Insights (Execution)¶
Cognitive mode: Deep analytical synthesis. Capture human-sourced domain knowledge observed during execution, then ADD implications — what does this insight mean for the project?
Human-Only Test: Would this insight be unknown without the user saying it? If an agent can discover it by reading code — it's NOT a strategic insight, it's a Fact Candidate (§6).
When to fill: Only when the human provides domain knowledge, corrections, or strategic context DURING execution. If no human interaction occurred — write "No strategic insights."
Categories: conventions.md §10.1.
| # | Insight | Category | Source |
|---|---|---|---|
| S1 | {insight} | {category — see §10.1} | User, |
Source format: Use reference patterns (e.g.,
[RF TFW-18](../../tasks/TFW-18__knowledge_consolidation/RF__PhaseB__knowledge_quality.md),D24). See compilable_contract.md §2.
8. Diagrams¶
Cognitive mode: Technical engineering documentation. Visualize architecture, data flow, component interaction, or sequence diagrams for the work completed in this phase.
Formats: ASCII, mermaid, or structured tables. Focus: HOW the system is built — components, layers, protocols, data flow.
If no diagrams are relevant — write "No diagrams."
RF — {PREFIX}-{N} / Phase {X}: {Title} | YYYY-MM-DD