Platform Rubric Manager

Platform Rubric Manager

Use this skill when platform-level judgment standards need to be created, reviewed, compared, or updated.

Use This For

• creating one rubric per platform
• reviewing the current rubric before scoring content
• updating a rubric after repeated retro signals
• comparing old and new rubric versions

Do Not Use This For

• writing the source article
• rewriting platform drafts
• scoring a specific draft directly
• publishing content

Inputs

• platform name
• optional existing rubric file
• optional retro findings
• optional benchmark examples

Outputs

• platform rubric draft or revision
• concise change summary
• explicit notes on what was added, removed, or tightened

Recommended File Layout

rubrics/
  wechat.yaml
  xiaohongshu.yaml
  x.yaml
  tiktok.yaml
  facebook.yaml
  reddit.yaml
  devto.yaml

Rubric Structure

Each platform rubric should define at minimum:

• platform goal
• audience expectation
• scoring dimensions
• strong signals
• risk signals
• tone rules
• CTA rules when relevant

Example shape:

platform: wechat
goal: build authority and complete an argument
dimensions:
  - title_strength
  - opening_grab
  - structure_depth
  - information_density
  - insight_strength
strong_signals:
  - result-first opening
  - clear section rhythm
  - concrete examples supporting a judgment
risk_signals:
  - opening too slow
  - generic conclusion
  - weak title promise

Modes

1. View

Use when a platform draft is about to be scored and the current rubric must be checked first.

2. Update

Use when retro results show repeated misses or repeated wins that should change future judgment.

3. Diff

Use when comparing an old rubric and a proposed new rubric.

4. Promote

Use when a revised rubric is judged better and should replace the working version.

Core Pattern

1. Keep one rubric per platform.
2. Do not merge all platforms into one universal scoring file.
3. Update a rubric only after repeated evidence, not one random outlier.
4. Prefer small explicit changes over full rewrites.

Common Mistakes

• Using one rubric across all platforms.
• Updating the rubric after a single surprising result.
• Letting tone rules stay vague.
• Keeping old dimensions that no longer explain results.

Quick Rules

• Platform logic is local, not universal.
• Retro data should refine the rubric, not replace judgment.
• Small repeated misses matter more than one lucky hit.
• Rubrics should become sharper over time, not longer by default.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: GiantClam
• Source: GiantClam/auto-viral-article-writer
• License: MIT

Install and usage instructions live in the source repository linked above.