opportunity-judgement

name	opportunity_judgement
description	Judge whether the old content deserves a ReCast task, no-action report, or further observation.
stage	decision

Skill: opportunity_judgement

Mission

Decide whether to generate a ReCast task.

Decision Options

• generate task card
• generate no-action report
• observe only

First-Level Decision: Yes or No

Yes means a real ReCast signal exists and the system should generate a task card.

No means there may be an ordinary topic direction, but there is not enough old-content ReCast value to justify a task card.

Useful topic direction is not the same as ReCast signal.

ReCast Signal Standards

A ReCast signal can come from:

• a post-publication change that directly affects the old claim, comparison, tutorial path, recommendation, or test frame
• a clear audience signal tied to the old content's claim or reusable frame
• a strong old content asset with enough engagement or comment demand to justify reuse
• a verifiable drift in the old content's external dependency

For change/update/drift signals:

• The baseline is the old post's posted_at.
• Evidence before posted_at can be background context only.
• Evidence before posted_at must not be used as a post-publication change trigger.
• If evidence synthesis marks occurred_after_post=false, that evidence cannot increase the post-publication change signal.
• If the old post did not specify a model/version, do not infer a hidden version assumption. Treat the version as unknown.

Confidence Rules

High: a real ReCast signal exists, the expected ROI is high, and the task is immediately worth producing. Medium: a real ReCast signal exists, but one part of asset strength, audience signal, evidence strength, or scope control is imperfect. Low: a real ReCast signal exists, but it is weak. A task card may be useful as a lightweight option, not as an urgent production recommendation. None: no meaningful ReCast signal. Return no-action.

Rules

• Do not force action to make the product look active.
• A medium-confidence retest is valid if framed honestly.
• If no action is recommended, produce a no-action report instead of empty output.
• Do not upgrade weak engagement and isolated comments into a strong signal without explaining why they are enough.
• Do not treat industry activity or generic product news as ReCast evidence unless it directly affects the old content.
• If the evidence only supports a normal content idea, set should_generate_task to false.
• Do not claim the user used GPT-4o, GPT-5, or any specific model unless the old post explicitly says so.

Output Contribution

Contribute action decision, ReCast signal existence, signal strength, confidence, reasons, and caveats.