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 Admin agent is the final stage of the Polysona content loop. It takes the draft selected from the virtual-follower’s TOP 5, saves it to content/published/, and generates a platform-specific publishing checklist so you know exactly what to do before hitting post. It also captures structured performance metadata — impressions, likes, reposts, replies, and engagement rate — so that real engagement data can be fed back into the persona over time. This feedback loop means the pipeline improves with every published piece.

​
Configuration

FieldValue
nameadmin
toolsRead, Write, Bash
Codex command$publish
Claude Code command/publish

​
Role

The Admin agent performs four jobs:
  1. Publishing storage — saves the final selected draft to content/published/ with a standardized filename.
  2. Platform-specific checklist — returns actionable publishing steps appropriate to the target platform.
  3. Metadata capture — records platform, published_at, hook, engagement_target, and actual_engagement fields alongside the draft.
  4. Feedback loop updates — when engagement data is later added, updates nuance.md platform patterns based on what worked and what failed.

​
Mandatory Execution Workflow

1

Parse platform and draft

Read the target platform and the final draft text from the current request.
2

Derive a filesystem-safe slug

Generate a slug from the draft topic or its central phrase to use as part of the output filename.
3

Write published file (required)

Use the Write tool to create the published file at:
content/published/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 file to confirm it exists and verify the final contents. Only after successful verification, return the saved path and the publishing checklist.
5

Return checklist and path

Provide the confirmed saved path and the platform-specific publishing checklist.
6

Report on failure

If the write fails, report it explicitly. Do not pretend the file was saved.

​
Metadata Contract

Every published file includes a structured metadata block with the following fields:
FieldDescription
platformTarget publishing platform
published_atISO date of publication
statusready-to-post at creation time
personaPersona name or ID
char_countCharacter count of the final draft
hookThe opening line (for later pattern analysis)
engagement_targetAnticipated engagement goal
actual_engagementInitially blank — filled in after posting

​
Published File Template

# Published: <PLATFORM> — <SLUG>

## metadata

platform     | <platform>
published_at | <YYYY-MM-DD>
status       | ready-to-post
persona      | <persona-name-or-id>
char_count   | <character-count>

## final_draft

<final draft text>

## performance_tracking

metric           | baseline | day1 | day7 | day30
impressions      | —        | —    | —    | —
likes            | —        | —    | —    | —
reposts          | —        | —    | —    | —
replies          | —        | —    | —    | —
engagement_rate  | —        | —    | —    | —

​
The Feedback Loop

The performance_tracking table is designed to be filled in after the content goes live. When you update the day1, day7, or day30 columns with real engagement data, the Admin agent can read that data and update nuance.md platform patterns accordingly — capturing what hook structures performed well, which CTA types drove replies, and which topics over- or under-delivered against the engagement target.
This feedback loop is what makes Polysona’s persona model self-improving. Each published piece, once its performance data is captured, refines the voice and platform patterns the content-writer will use in the next cycle.

​
Integration Rule

All published outputs must land in content/published/. This is a hard requirement — the dashboard and content pipeline discover published content by scanning this directory. Platform-specific subdirectories or persona-scoped paths are not used.
Do not save published outputs to persona-specific subdirectories (e.g., personas/{id}/published/). The pipeline’s discovery logic only reads from content/published/.

​
Operating Guardrails

  • Preserve append-first history for published outputs. Never overwrite an existing published file.
  • Keep metadata machine-readable and human-auditable.
  • The actual_engagement field starts blank — fill it only after real data is available.

​
Invocation

$publish

Build docs developers (and LLMs) love

Get started for freeTalk to us