Prompt Composition
1 requirement
·
5 scenarios
Prior AI Email Response Tagging
JR-prompt-composition-001
The system SHALL detect and tag prior AI email responses in quoted email threads before rendering the system prompt. AI responses are detected by:
- Start: Lines matching
- End: Detected using fallback markers in priority order:
1.
2. Triple ZWSP (
3. Workflow emojis:
4.
AI response content after the detected end marker SHALL be truncated and replaced with a summary note referencing
Each detected AI response SHALL be wrapped in semantic tags:
- Opening:
- Closing:
- Start: Lines matching
From:.*<email> where email is in DEFAULT_BLOCKED_EMAILS- End: Detected using fallback markers in priority order:
1.
-Junior or - Junior signature (primary - keeps signature, strips after)2. Triple ZWSP (
\u200B\u200B\u200B) - progress block indicator3. Workflow emojis:
🔧, 🕒, ⚙️ - tool output indicators4.
[[ticket_...]] marker - final boundary (never included)AI response content after the detected end marker SHALL be truncated and replaced with a summary note referencing
[prior_analyses].Each detected AI response SHALL be wrapped in semantic tags:
- Opening:
[PRIOR_EMAIL_RESPONSE_BY_AI]- Closing:
[/PRIOR_EMAIL_RESPONSE_BY_AI]
5 test scenarios
- AI response with signature JR-prompt-composition-001.1
- Interim AI response without signature JR-prompt-composition-001.2
- Multiple AI responses in thread JR-prompt-composition-001.3
- User-only thread JR-prompt-composition-001.4
- AI response detection from blocked emails list JR-prompt-composition-001.5