Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/LilMGenius/polysona/llms.txt

Use this file to discover all available pages before exploring further.

The Virtual-Follower is Polysona’s quality assurance layer. Before any draft goes to the admin for publishing, it passes through a simulation of 20 distinct audience archetypes — each reading the draft through their own lens and scoring it independently. The agent then performs a rolemodel gap analysis, comparing draft style against the benchmarks in accounts.md, and returns a ranked TOP 5 recommendation list. The user selects one of the TOP 5 before the admin agent publishes.

​
Configuration

FieldValue
namevirtual-follower
contextfork
toolsRead, Write
Codex command$qa
Claude Code command/qa

​
Role

The Virtual-Follower performs three jobs:
  1. Multi-audience simulation — evaluates each draft as 20 different follower archetypes, applying each archetype’s occupation, age, and platform expectations.
  2. Rolemodel-gap scoring — compares draft style against top-performing cues in accounts.md rolemodel entries, identifying both alignment and deficiency.
  3. TOP 5 recommendation — aggregates scores and returns a numbered TOP 5 list with rationale for each pick.

​
The context: fork Execution Mode

The Virtual-Follower runs in context: fork, which creates an execution context that is fully isolated from the content-writer’s generation session. This is a deliberate architectural guarantee: QA evaluation must reflect an independent reader’s reaction, not a continuation of the drafting context.
Do not inject content-writer context into the Virtual-Follower session manually. The context: fork boundary ensures that the 20 archetypes evaluate the draft as written — not as the author intended it to be understood.

​
The 20 Follower Archetypes

Scores are computed across all 20 archetypes from FOLLOWER_ARCHETYPES in the API. The full archetype table:
IDLabelAgeGenderOccupation
f0120s female office worker24Foffice
f0230s male developer32Mdev
f0340s self-employed43Mself
f04startup founder29Mfounder
f05general follower27Fgeneral
f0620s male student21Mstudent
f0730s female marketer35Fmarketing
f0840s male executive47Mexecutive
f0920s female creator23Fcreator
f1030s male engineer31Mengineer
f1150s female entrepreneur52Fentrepreneur
f1220s non-binary designer26NBdesign
f1330s male product manager34Mpm
f1440s female educator41Feducation
f1520s male gamer22Mgaming
f1630s female researcher33Fresearch
f1740s male consultant44Mconsulting
f1820s female influencer25Finfluencer
f1930s male journalist37Mmedia
f2050s male investor55Minvestor

​
The 5 Scoring Dimensions

Each archetype scores each draft on five dimensions. All scores fall in the 40–95 range.
DimensionKey Question
hookDoes the first line stop scrolling?
empathyDoes the intended audience relate quickly?
shareWould they RT / repost / share?
ctaWould they comment, follow, or click?
platform_fitDoes it match the platform’s reward patterns?
Each archetype’s total score is the sum of its five dimension scores. The Virtual-Follower aggregates all archetype scores per draft to identify the strongest-performing variations. The TOP 5 are the five recommended draft variations with the highest aggregate scores across all archetypes.

​
TOP 5 Selection and Output

The Virtual-Follower returns a numbered TOP 5 list. For each of the five top-ranked recommendations, the output includes:
  • Total score — sum of all dimension scores
  • Strengths — what the draft does well for this archetype
  • Weaknesses — what is missing or underperforming
  • Rolemodel gap note — where the draft aligns with or diverges from accounts.md rolemodel style benchmarks

​
Input Sources

SourceUsed For
content/drafts/ (latest file)Draft text to evaluate
accounts.md (rolemodel section)Rolemodel style benchmark comparison
accounts.md (virtual section)Audience archetype context
If no draft file exists in content/drafts/, the Virtual-Follower stops and reports that QA is blocked until /content creates one. It does not proceed on draft-less input.

​
Mandatory Execution Workflow

1

Read the latest draft

Load the most recent file from content/drafts/ unless the user explicitly provided draft text in the current request.
2

Evaluate across archetypes and dimensions

Score the draft against all 20 follower profiles on all 5 dimensions. Run rolemodel gap analysis against accounts.md.
3

Write QA report (required)

Use the Write tool to save the QA report to:
content/qa/YYYY-MM-DD-platform-slug.md
This step is mandatory. Do not proceed without writing the file.
4

Verify the written file

Immediately use the Read tool on the saved QA report to confirm it exists and reflects the evaluation. Only after successful verification, return the TOP 5 and the confirmed saved path.
5

Report on failure

If the write fails, report it explicitly. Do not pretend QA storage succeeded.

​
QA Report File Template

# QA Report — <PLATFORM> — <SLUG>

## metadata

evaluated_at | <YYYY-MM-DD>
persona      | <persona-id>
source_draft | <path>

## top5

1. <recommendation>
   - score: <total>
   - strengths: <text>
   - weaknesses: <text>
   - rolemodel_gap: <alignment and deficiency notes>

2. <recommendation>
   ...

​
Invocation

$qa
After reviewing the TOP 5, select the draft variation you want to publish and pass it to the admin agent with $publish / /publish.

Build docs developers (and LLMs) love

Get started for freeTalk to us