domains / families / research / result-to-claim

result-to-claim

combined

Use when experiments complete to judge what claims the results support, what they don't, and what evidence is still missing. Codex MCP evaluates results against intended claims and routes to next action (pivot, supplement, or confirm). Use after experiments finish — before writing the paper or running ablations.

When to use

• After a set of experiments completes (main results, not just sanity checks)
• Before committing to claims in a paper or review response
• When results are ambiguous and you need an objective second opinion

When not to use

• writing the paper
• running ablations
• sanity checks
• scheduling via /loop, /schedule, or CronCreate

Skill page

Conflict

The two copies share the same core gate (collect results, Codex judgment, route on verdict), but the wanshuiyin copy adds a mandatory deterministic evidence pre-check (Step 1.5) that mechanically verifies cited numbers before the Codex call, plus a scheduling warning and a legacy path for research_contract.md. The brycewang copy contains the full Step 2 Codex prompt, Step 3 normalization, and Step 4 routing rules that the wanshuiyin copy truncates. Neither copy is a strict superset of the other, so crowning one would hide the other's material steps.