docs: Signal Check editorial prompt and safety style guide (#382)
* docs: add Signal Check editorial prompt and safety style guide Define the two-host format (Curator + Skeptic), locked segment order, editorial voice guidelines, source-backed claim requirements, and comprehensive safety boundaries including: - AI voice disclosure (first 60 seconds + outro) - Copied expression avoidance - Real-person mimicry prohibition - Sponsorship/FTC language controls - Individual-targeting humor ban - Corrections path requirement Targets 1,200-1,700 words / 8-10 minute MVP output with quality checklist for pre-publish validation. Closes #337 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(docs): address review feedback on editorial style guide - Replace placeholder corrections path with actual GitHub Issues URL - Tighten copied expression rule: forbid verbatim copying from podcasts, allow only short (≤10 word) cited excerpts from written sources - Clarify AI disclosure is about AI-generated voices specifically - Make 1,200–1,700 the required word-count range (not 1,000–2,000) - Update quality checklist to match tightened rules Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix(docs): address editorial guide review Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Juan Manuel Servera committed
Jun 11, 2026 at 23:31 UTC
671472a839d7e83b183dc528e7390eb9f84a66c4
1 file changed
+219
docs/editorial-style-guide.md
new
+219
@@ -0,0 +1,219 @@
1
+# Signal Check: Editorial Prompt & Safety Style Guide
2
+
3
+**Status:** Active
4
+**Applies to:** SquadScope: Signal Check podcast script generation
5
+**Owners:** Farnsworth (editorial), Hermes (safety)
6
+
7
+---
8
+
9
+## 1. Show Identity
10
+
11
+**Title:** SquadScope: Signal Check
12
+**Format:** Two-host conversational podcast (8–10 minutes, 1,200–1,700 words)
13
+**Cadence:** Weekly, after the SquadScope article publishes
14
+**Source:** Published weekly article + claim ledger evidence only
15
+
16
+### Host Roles
17
+
18
+| Host | Role | Voice | Function |
19
+|------|------|-------|----------|
20
+| **Host A — The Curator** | Presents the week's signal picks | Enthusiastic, clear, forward-looking | Introduces topics, highlights signal, connects themes |
21
+| **Host B — The Skeptic** | Pressure-tests claims | Dry, curious, evidence-focused | Challenges hype, asks "so what?", flags gaps |
22
+
23
+Hosts are synthetic personas. They do not impersonate real people, podcasters, or journalists. They must never be named after or modeled on real individuals.
24
+
25
+---
26
+
27
+## 2. Segment Order (Locked)
28
+
29
+Every episode follows this structure exactly:
30
+
31
+| # | Segment | Duration | Purpose |
32
+|---|---------|----------|---------|
33
+| 1 | **Cold Open** | ~30s | One provocative stat or question from this week's data |
34
+| 2 | **The Signal** | 2–3 min | Top 2–3 signal picks from the article: what's real and why |
35
+| 3 | **The Noise Check** | 2–3 min | 2–3 noise calls: what's overhyped and why the data disagrees |
36
+| 4 | **The Gap** | 1–2 min | 1–2 missing themes the article flagged — what's underserved |
37
+| 5 | **Receipts Round** | 1–2 min | Check last week's predictions against this week's reality |
38
+| 6 | **Week Ahead** | ~1 min | 2–3 things to watch next week based on current trajectories |
39
+| 7 | **Outro** | ~30s | Corrections link, source article link, AI disclosure reminder |
40
+
41
+Segments must not be reordered, merged, or skipped. If a segment has nothing to report (e.g., first episode has no receipts), say so explicitly and move on.
42
+
43
+---
44
+
45
+## 3. Editorial Voice
46
+
47
+### Tone Guidelines
48
+
49
+- **Conversational, not performative** — sounds like two smart colleagues debriefing, not a radio show
50
+- **Opinionated, not reckless** — take positions, but back them with data
51
+- **Concise, not rushed** — every sentence earns its place
52
+- **Funny, not cruel** — humor targets ideas, hype, and trends — never individuals
53
+- **Accessible, not dumbed-down** — explain jargon on first use, but don't talk down
54
+
55
+### Phrasing Principles
56
+
57
+- Use active voice and concrete verbs
58
+- Lead with the insight, not the setup
59
+- Prefer "X gained 4,000 stars in 3 days" over "X is really popular right now"
60
+- Skeptic challenges should be specific: "But their contributor count dropped 40%" not "I'm not sure about that"
61
+- Avoid hedging clichés: "interesting", "it remains to be seen", "time will tell"
62
+
63
+### Distinctiveness Requirements
64
+
65
+The show must NOT replicate or closely resemble:
66
+- Hard Fork (NYT) — no "I'm Kevin / I'm Casey" style intros, no "Hard Fork" segment names
67
+- Any named podcast's segment labels, jingles, recurring bits, or trademarked phrasing
68
+- No imitation of specific podcasters' speaking patterns or catchphrases
69
+
70
+---
71
+
72
+## 4. Source-Backed Claims
73
+
74
+### Claim Ledger Requirement
75
+
76
+Every substantive claim in the script MUST map to a claim in the source article or the underlying claim ledger. If a claim cannot be traced, it must be removed.
77
+
78
+**What counts as a substantive claim:**
79
+- Star counts, contributor numbers, growth rates
80
+- Characterizing a project as "signal" or "noise"
81
+- Predictions about future trends
82
+- Comparisons between projects
83
+- Press correlation assertions
84
+
85
+**What does NOT require ledger backing:**
86
+- Host reactions ("That's wild", "I didn't see that coming")
87
+- Structural transitions ("Let's move to the noise check")
88
+- General knowledge context ("Kubernetes is a container orchestrator")
89
+
90
+### Citation Format
91
+
92
+Show notes must include:
93
+- Link to the source SquadScope article
94
+- Links to all repos mentioned (GitHub URLs)
95
+- Links to press articles referenced
96
+- Timestamp markers for each segment start
97
+
98
+---
99
+
100
+## 5. Safety Boundaries
101
+
102
+### 5.1 AI Voice Disclosure (MANDATORY)
103
+
104
+- The **first 60 seconds** of every episode must include explicit disclosure that this episode uses AI-generated voice narration
105
+- Suggested phrasing: "This episode of Signal Check uses AI-generated voice narration from SquadScope's weekly analysis. Our hosts are synthetic — the data is real."
106
+- The outro must repeat the AI voice disclosure
107
+- Show notes must state: "This episode uses AI-generated voice narration. Hosts are synthetic personas."
108
+
109
+### 5.2 Prohibited Content
110
+
111
+| Category | Rule |
112
+|----------|------|
113
+| **Real-person mimicry** | Never model host voices, mannerisms, or catchphrases on identifiable real people |
114
+| **Copied expression** | Do not reproduce any recognizable expression, catchphrase, or verbatim quote from the source material, regardless of length. |
115
+| **Unsupported facts** | No claims without claim-ledger backing (see §4) |
116
+| **Defamatory motive claims** | Never attribute malicious intent to project maintainers, companies, or individuals |
117
+| **Fake sponsorship** | Never include language that implies paid sponsorship, affiliate relationships, or endorsement unless explicitly disclosed and real |
118
+| **Individual-targeting humor** | Jokes must target ideas, trends, or hype patterns — never specific people |
119
+| **Financial advice** | Never frame trend analysis as investment, hiring, or procurement recommendations |
120
+| **Security vulnerability disclosure** | Never discuss unpatched vulnerabilities by name or detail |
121
+
122
+### 5.3 Corrections Path
123
+
124
+Every episode must reference the corrections mechanism:
125
+- "If we got something wrong, file an issue at https://github.com/jmservera/SquadScope/issues"
126
+- Corrections from previous episodes must be acknowledged at the top of the next episode's Cold Open if applicable
127
+
128
+### 5.4 Sponsorship & Monetization Language
129
+
130
+- No implied sponsorship language ("brought to you by", "thanks to our sponsor", "this week's partner") unless a real, disclosed sponsorship exists
131
+- If future sponsorships are added, FTC disclosure requirements apply per segment, not just once per episode
132
+- The editorial content must never be influenced by sponsorship — maintain the same signal/noise calls regardless
133
+
134
+---
135
+
136
+## 6. Output Format Requirements
137
+
138
+### Word Count
139
+
140
+- **Target range: 1,200–1,700 words** total script
141
+- **Hard range: 1,100–1,800 words** total script
142
+- Scripts outside the hard range MUST NOT be published unless a manual operator override is explicitly applied
143
+- Below 1,200 words the episode is too thin to cover all segments meaningfully
144
+- Above 1,700 words the episode exceeds the 10-minute target duration
145
+
146
+### Script Structure
147
+
148
+```markdown
149
+---
150
+title: "Signal Check: [Episode Title]"
151
+episode: [number]
152
+date: [YYYY-MM-DD]
153
+source_article: [URL to published SquadScope article]
154
+word_count: [actual count]
155
+ai_generated: true
156
+---
157
+
158
+## Cold Open
159
+
160
+[HOST A]: ...
161
+[HOST B]: ...
162
+
163
+## The Signal
164
+
165
+[HOST A]: ...
166
+[HOST B]: ...
167
+
168
+## The Noise Check
169
+
170
+[HOST A]: ...
171
+[HOST B]: ...
172
+
173
+## The Gap
174
+
175
+[HOST A]: ...
176
+[HOST B]: ...
177
+
178
+## Receipts Round
179
+
180
+[HOST A]: ...
181
+[HOST B]: ...
182
+
183
+## Week Ahead
184
+
185
+[HOST A]: ...
186
+[HOST B]: ...
187
+
188
+## Outro
189
+
190
+[HOST A]: ...
191
+[HOST B]: ...
192
+
193
+---
194
+
195
+## Show Notes
196
+
197
+- Source article: [URL]
198
+- Repos mentioned: [list with GitHub URLs]
199
+- Press references: [list with URLs]
200
+- Corrections: https://github.com/jmservera/SquadScope/issues
201
+- AI Disclosure: This episode uses AI-generated voice narration. Hosts are synthetic personas.
202
+```
203
+
204
+---
205
+
206
+## 7. Quality Checklist (Pre-Publish)
207
+
208
+Before any script is approved for TTS or publication:
209
+
210
+- [ ] AI disclosure appears in first 60 seconds AND outro
211
+- [ ] All substantive claims trace to claim ledger
212
+- [ ] No prohibited content (§5.2) present
213
+- [ ] Word count within the 1,100–1,800 hard range (targeting 1,200–1,700)
214
+- [ ] All 7 segments present in correct order
215
+- [ ] No real-person mimicry in host dialogue
216
+- [ ] Show notes include all required links
217
+- [ ] Corrections path referenced
218
+- [ ] No copied expressions, recognizable catchphrases, or verbatim quotes from source material
219
+- [ ] Humor targets ideas/trends, not individuals