public-opinion-scenario

name	public-opinion-scenario
description	Scenario pack for Chinese public-opinion risk analysis under learn.skill. Use when the orchestration skill needs domain-specific rules, eval sets, and memory for event-pack based public-opinion assessment.

public-opinion scenario

This scenario pack defines the first domain for learn.skill: Chinese public-opinion risk analysis at general-practitioner level.

Scope

• assess event-pack risk, not single-post moderation only
• separate verified fact from ambient online signal
• produce structured outputs for downstream judging
• support abstention when evidence is weak or contradictory

Read order

1. scenario.yaml
2. memory.md
3. skill_package/
4. eval/ only according to the current stage and permission boundary

Boundary reminders

• Executors do not see corpus/, memory.md, or judge bundles.
• Final hidden materials are judge-only unless the scenario manifest says otherwise.
• Repairs may use redacted summaries from canary_hidden and aggregate outcomes from final_hidden.