Back to directory

fablize

A Claude Code plugin that adds Fable's verified working procedures, such as evidence-gated completion and systematic investigation, to Opus and other Claude models.

fablize

fablize is a Claude Code plugin developed after a controlled comparison between Fable 5 and Opus 4.8 across roughly 1,500 tool calls. The comparison found that on closed, answer-bearing work the two models performed about the same, and that the gap on open-ended work came from model capability rather than procedure.

The plugin applies only the procedures the comparison verified as transferable: grounding completion in actually running and observing the artifact, a multi-story completion gate that requires evidence before work counts as done, systematic investigation that moves from reproduction to hypotheses to a causal chain, and a deterministic hook that catches early stopping. Traits the comparison found to be model capability, such as open-ended creative judgment, are left alone; fablize tells the agent to escalate those instead of pretending to have them.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.