Appendix E — Templates

Six artifacts, each a page you fill in rather than read: the specification from Part II, the prompt structure from 5.2, the media kit from 9.5, the disclosure duties from 11.3, the single-post procedure from 7.6, and the records from 7.4 and 11.5. Keep the shape; adapt the fields.

Template 1 — Character bible form

Two files (2.3): a machine-readable core the pipeline reads, and a prose file for the parts only a person reads. Leave a field blank rather than inventing a detail you will not maintain.

Part A — machine-readable core

persona:            <name, exact spelling>
pronunciation:      <phonetic reference or audio file>
version:            <date + reason; never overwritten>
trigger_token:      <adapter trigger from 4.5>
palette:            <three dominant colors, two accents>
prohibited_words:   [<list>]
canon:              [<ten falsifiable facts>]
recurring_names:    [<named characters and their frequency cap>]
platforms:          [<handle per platform, identical spelling>]
disclosure_line:    <verbatim, from 11.3>

Part B — identity

• Name, age at launch, and whether age advances:
• Origin, city, languages, primary language:
• Education and occupation:
• Relationship status:
• Family and recurring friends:
• The one-sentence reason to follow (2.2):

Part C — physical canon

• Height and build, with the dataset's head-to-body ratio (4.4):
• Distinguishing mark — the QA checkpoint:
• Hair: length, color, texture, styles worn:
• Face: the twelve attributes on the reference plate (5.1):
• Hands: nails, jewelry, rings — the most reliable tell (1.4):
• Voice: register, pace, accent, verbal tics:

Part D — voice rules at sentence level

• Average sentence length and rhythm:
• Vocabulary used; words never used, as a list:
• Register, constant across platforms:
• Emoji and punctuation conventions:
• Five opinions, specific enough to be arguable:
• When wrong: correction in character within a week, post never deleted (2.3):

Part E — boundaries: what the persona never says or does

• Subjects refused, with the in-character reason:
• Claims never made — medical, legal, or financial advice stated as fact:
• Depictions never produced (12.4):
• The one thing the persona is bad at on purpose, and how often it recurs:
• Presented as a real human being? Never (9.5's clause to refuse):

Part F — timeline

• Persona birth date and account launch date:
• Dated entries: what changed, when, what published under the previous version:
• Canon life events, with dates:
• Style changes, with first-publication dates:
• Decision log: question, answer, date:

Part G — platform presence

• Platform, handle, and the format it carries:
• Bio per platform, disclosure line first:
• Pinned posts and their job:
• The destination each profile routes to (10.1):
• Which accounts carry paid work, and the label regime (11.3):

The canon test, before every publication: does this asset contradict any line in Part A or Part E? If yes, it does not publish.

Template 2 — Prompt templating file, 40 slots

Six parts, in order (5.2): identity (fixed), scene (the work), light (one per prompt), lens and framing (frozen vocabulary), style and grade (photographic register only, since the grade is applied in post), and quality and texture (the retention phrases).

[identity]  <trigger token>, <adapter and reference conditioning>, <age>
[scene]     <location>, <time of day>, <action>, <wardrobe>, <two or three props>
[light]     <one condition from the kit's three>
[frame]     <focal look>, <shot size>, <camera height>
[style]     <photographic register>
[quality]   preserve facial identity, realistic skin texture, natural skin pores
[negative]  blurry, jpeg artifacts, extra fingers, malformed hands, asymmetric eyes,
            plastic skin, doll-like, watermark, text, overexposed

Forty variants, roughly two weeks at three posts a day. Refill quarterly from the review slot (2.6), with the wardrobe and prop words drawn from your prop closet (2.3).

A filled variant, assembled

[identity]  <trigger token>, <adapter and reference conditioning>, woman, mid-20s
[scene]     standing in a small kitchen, morning, holding a mug,
            stone counter, a plant and a laptop on the counter
[light]     soft window daylight from the left, overcast, no harsh shadow
[frame]     50mm-equivalent look, medium shot, eye-level camera
[style]     documentary photography, available light, unretouched
[quality]   preserve facial identity, realistic skin texture, natural skin pores
[negative]  blurry, extra fingers, malformed hands, asymmetric eyes,
            plastic skin, doll-like, watermark, text, overexposed

Three rules carry it. The scene section holds roughly half the prompt, and no facial description appears, because a written face competes with the reference layer. One light only, since two sources produce the impossible-shadow tell (1.4). The grade is absent, since post applies it to every frame.

Template 3 — Media kit for a synthetic persona

One page plus one sample folder. The page answers questions; the folder proves quality. A brand buys a deliverable somebody inside the company can defend (9.5).

The page

• Persona: name, niche, and the disclosure line verbatim — __________ is an AI-generated persona. Images and voice are synthetic.
• Audience snapshot: followers per platform, engagement rate, free-to-paid split, three sentences on who they are.
• Content formats: 4:5 stills, 9:16 clips, carousels, voiceover pieces, with quantities per campaign.
• Reach: trailing 30-day average views, saves, shares. Median month, not best (9.2).
• Capacity: turns per weekday, turnaround, QA survival rate from 5.6. No human creator's kit matches this line.
• Deliverables: formats, variations per campaign, revision cap, in writing.
• Rights offered: channels, window, territory, amplification included or not.
• Compliance: disclosure placement (11.3), provenance records (11.5), asset metadata (7.4).
• Likeness: an owned asset, provenance documented, no real person referenced (11.4).
• Contact: one person, one address, one ask.

The rate card

Category choice moves the rate more than follower count does, and the lower band prices the risk of an unknown account (9.5).

Template 4 — AI UGC agreement: clauses to draft