Skip to content
SkillModel-led777 tokens in use

False-Positive Reviewer

Interprets AI-writing flags with care: what the evidence can and cannot establish, the human reasons a flag appears, and which evidence helps a real decision.

5,770 stars on the repository MIT Licenseconorbronsdon/avoid-ai-writing · skills/false-positive-reviewer
/plugin install avoid-ai-writing@conorbronsdon-skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing False-Positive Reviewer costs

49tokens idle
777tokens in use

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Fair
  • Frontmatter15/15
  • Structure23/25
  • Examples0/15
  • Output10/15
  • Size10/15
  • Resources0/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Excellent
  • Development activity96
  • Maintenance100
  • Community60
  • Maturity32
  • Momentum40

False-Positive Reviewer answers questions such as does this detector score prove the text was written by AI. It separates deterministic detector hits, model-only observations and facts the user supplied, explains plausible human causes for each signal, and declines to turn a score into a verdict about cheating, fraud or hiring suitability.

It suggests more useful evidence for consequential decisions, such as drafts, revision history and a conversation with the writer.

What You Get

  • A breakdown of which observations came from executed tooling, model-only review or user-supplied context
  • An explanation of why each signal may have a human cause
  • A statement of what the text shows, what it may suggest and what it cannot establish
  • A list of more probative evidence, such as source history, drafts and revision logs
  • A return-to-router request when new signal collection or a rewrite is wanted

Common Use Cases

  • A teacher asks whether a high detector score shows a student used AI
  • A hiring manager wants to know how much weight a flagged cover letter deserves
  • An editor questions why a human-written article trips many flags
  • A writer working in a second language wants to understand false positives
  • A team writes a policy on how detector results may be used

What Makes It Unique

The skill is built to resist the pull toward a verdict. It keeps score and label as noisy evidence, refuses to invent confidence values or a probability of authorship, and points to process evidence for decisions with consequences. Its place in the skill graph is terminal, which prevents a loop between reviewer and detector.

Under The Hood

This is an instructions-only skill with no scripts. It loads one instruction file and reads the rulebook and routing files of the sibling skills as references. It does not run commands, call APIs or write files. When more evidence is needed it hands control back to the router, which may run the detector.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers