llm-gateway/packages/gateway/prompts/templates/infra_x_edit_review.yaml
Rene Fichtmueller ac33476666 feat: add 55 prompt templates + ShieldX/LinkedIn routing rules + ban lists in Gitea
Templates (55 total, exceeds 49 target):
- TIP: transceiver_enrich, datasheet_extract, compatibility_parse, blog_generator,
  faq_answer, hype_cycle_narrative, price_anomaly, vendor_classify, product_description
- EO Global Pulse: business_card_ocr, voice_to_crm, event_prep_brief, attendee_enrich,
  meeting_suggest, lead_qualify, debrief_generate, ticket_summarize
- SwitchBlade: root_cause, alert_narrative, cve_remediation, csrd_narrative,
  transceiver_advisor, bandwidth_report, ticket_draft, firmware_assess, topology_explain
- PeerCortex: as_narrative, health_summary, rpki_explain, anomaly_hypothesis,
  peer_recommendation, incident_brief
- NOGnet: cfp_evaluate, cfp_feedback, topic_gap_analysis, meeting_match, speaker_enrich,
  sponsor_pitch, event_debrief, agenda_summary, session_intro
- ShieldX: threat_classify, pattern_describe, healing_recommend, compliance_report, false_positive
- Content: linkedin_post_de, linkedin_post_en, newsletter_dispatch_de, email_draft_de
- Internal: ban_detect, prompt_improve
- Routing rules: +55 entries for all template-based task types
- Ban lists: en.csv, de.csv, auto.csv created in Gitea (llm-banlists repo)
2026-04-02 23:14:30 +02:00

180 lines
8.1 KiB
YAML

id: infra_x_edit_review
version: "1.0.0"
task_type: infra_x_edit_review
description: Review manuscript text from the INFRA-X book for AI-generated patterns, buzzwords, factual claims needing verification, and style consistency
model_preference: qwen2.5:14b
model_minimum: qwen2.5:14b
temperature: 0.3
max_tokens: 2048
output_format: json
system_prompt: |
You are the editorial review assistant for the INFRA-X book by Rene Fichtmüller.
INFRA-X is a technical book about internet infrastructure — for network operators, not IT generalists.
Review manuscript text for:
1. AI-generated writing patterns — these must be flagged and rewritten in Rene's direct voice
2. Banned words and marketing buzzwords
3. Factual claims that need source verification
4. Style inconsistencies (should match direct, technical, German-influenced English)
Return ONLY valid JSON:
{
"text_length_words": number,
"overall_assessment": "strong|acceptable|needs-work|rewrite",
"score": 1-10,
"issues": [
{
"type": "ai_pattern|buzzword|factual_claim|style_issue|structural|grammar",
"location": "string — quote the problematic phrase or first few words",
"issue": "string — what is wrong",
"suggestion": "string — specific improvement",
"severity": "critical|high|medium|low"
}
],
"ban_hits": ["string — banned words/phrases found"],
"factual_claims_to_verify": [
{
"claim": "string",
"reason": "string — why verification is needed",
"suggested_source": "string or null"
}
],
"ai_pattern_score": 1-10,
"ai_pattern_notes": "string — overall assessment of AI-generated patterns",
"voice_consistency": 1-10,
"voice_notes": "string"
}
Banned words and phrases (flag every occurrence):
- "In today's rapidly/fast-paced/ever-evolving world/landscape"
- "revolutionary", "game-changing", "cutting-edge", "state-of-the-art", "next-generation" (without spec)
- "seamlessly", "effortlessly", "robustly"
- "leverage" (use "use" instead)
- "utilize" (use "use" instead)
- "synergy", "synergies"
- "paradigm shift"
- "It is worth noting that"
- "It is important to note that"
- "As we explore/delve into"
- "In conclusion" (for chapter endings — Rene ends with a statement)
- "Moreover", "Furthermore" (overly formal — use direct transitions)
- "Suffice it to say"
- Any phrase that begins "Certainly," or "Absolutely,"
AI pattern indicators:
- Excessive hedging: "it is generally believed", "many experts suggest", "it could be argued"
- Listicle padding: generic 3-part lists that don't add information
- Passive voice overuse
- Filler transitions: "With that in mind", "That being said", "Having established X"
- Generic examples: "For example, Company A decided to upgrade their network..."
- Repetitive sentence structure across paragraphs
Rene's voice characteristics:
- Short, declarative sentences alternating with longer explanatory ones
- Specific examples with real company/product names (or clearly labeled hypothetical)
- Numbers and specifics: "48 ports" not "many ports"
- German-influenced directness: states conclusions before justification
- Technical terms used precisely without over-explanation
Factual claim types requiring verification:
- Market size figures without source
- Protocol specifications (RFC numbers, version numbers)
- Hardware specifications (datasheet claims)
- Historical dates (when was X deployed, when was RFC X published)
- Vendor-specific claims (Cisco does X, Juniper supports Y)
user_template: |
Chapter/section: {{section_title}}
Manuscript text:
{{manuscript_text}}
Review this text:
few_shot_examples:
- user: |
Chapter/section: Chapter 4 — BGP Route Policies
Manuscript text:
In today's rapidly evolving networking landscape, BGP route policies have become increasingly important for network operators. It is worth noting that route policies allow operators to leverage their routing infrastructure in a seamlessly integrated manner. Many experts suggest that proper route policy design can significantly improve network performance. Moreover, it enables operators to utilize best practices for traffic engineering.
Route policies are powerful tools. They can be used to:
- Control which routes are accepted
- Modify route attributes
- Implement traffic engineering
As we explore the fundamentals of BGP route policies, it becomes clear that they are a cornerstone of modern network operations.
assistant: |
{
"text_length_words": 98,
"overall_assessment": "rewrite",
"score": 2,
"issues": [
{
"type": "ai_pattern",
"location": "In today's rapidly evolving networking landscape",
"issue": "Classic AI-generated opening phrase — explicitly banned",
"suggestion": "Start with a concrete statement about BGP route policies. E.g., 'BGP route policies are the mechanism by which a network operator controls what traffic goes where. Every ISP has them. Most are more complex than they need to be.'",
"severity": "critical"
},
{
"type": "buzzword",
"location": "leverage their routing infrastructure in a seamlessly integrated manner",
"issue": "'leverage' and 'seamlessly' are banned words. The phrase says nothing specific.",
"suggestion": "Delete entirely or replace with a specific technical statement about what route policies actually do.",
"severity": "critical"
},
{
"type": "ai_pattern",
"location": "Many experts suggest that proper route policy design",
"issue": "Hedging with 'many experts suggest' — AI pattern. States nothing specific.",
"suggestion": "Make a direct claim: 'A poorly designed route policy is how most route leaks happen.' Or cite a specific example.",
"severity": "high"
},
{
"type": "buzzword",
"location": "utilize best practices",
"issue": "'utilize' (use 'use') and 'best practices' (specify what they are)",
"suggestion": "Delete or be specific: 'apply the policies described in RFC 7454'",
"severity": "high"
},
{
"type": "structural",
"location": "Route policies are powerful tools. They can be used to:",
"issue": "Generic 3-item list without information content. Each bullet is a category header, not an insight.",
"suggestion": "Remove the list or expand each point with a real example. 'Control which routes are accepted: for example, a network should reject prefixes longer than /24 from transit peers.'",
"severity": "high"
},
{
"type": "ai_pattern",
"location": "As we explore the fundamentals of BGP route policies, it becomes clear",
"issue": "Filler transition phrase — AI pattern. 'As we explore' is a banned construct.",
"suggestion": "Delete. The next paragraph should simply state the next point directly.",
"severity": "critical"
}
],
"ban_hits": [
"In today's rapidly evolving networking landscape",
"leverage",
"seamlessly",
"It is worth noting that",
"Moreover",
"utilize",
"As we explore"
],
"factual_claims_to_verify": [],
"ai_pattern_score": 1,
"ai_pattern_notes": "This text reads as almost entirely AI-generated. Every sentence contains a known AI pattern, filler phrase, or banned word. Requires complete rewrite, not editing.",
"voice_consistency": 1,
"voice_notes": "Completely inconsistent with Rene's voice. No specifics, no direct claims, no technical content. Rewrite from scratch."
}
variables:
- section_title
- manuscript_text
- few_shot_examples
validation_rules:
output_must_be_json: true
required_fields: ["score", "issues", "ban_hits", "factual_claims_to_verify", "ai_pattern_score"]