Result-to-Claim Gate
Judges whether experiment results support your intended claims and suggests the next step.
- Stars
- 18,482
- Licence
- MIT
- Files
- 1 file
What it does
After experiments finish, this skill makes the agent gather results from sources such as Weights & Biases runs, experiment log files, and server logs, then assess what the numbers actually support, what they don't, and what evidence is still missing. Before asking an outside reviewer model for a judgment, it runs a mechanical check that cited numbers really appear in the result files, which catches invented evidence. It then recommends whether to pivot, run supplementary experiments, or confirm the claim.
Great for
- Researchers who have finished a set of main experiments and want an independent check before writing a paper
- Teams preparing rebuttals or review responses that depend on whether results hold up
- Anyone who wants a structured second opinion when experiment results are ambiguous
Skip it if
- Routine sanity-check or debugging runs, since the skill is meant for main results and gating claims (not every run)
- Scheduled or repeated re-checks on a timer; the skill's own instructions say to run it once after results change, not on a loop
- Projects without a defined research claim, because the verdict depends on a stated intended claim
Try asking
- “Check whether my ablation results actually support the claim in my research contract before I write the paper”
- “Compare my latest W&B run against baselines and tell me what claims are still missing evidence”
- “Review EXPERIMENT_LOG.md and tell me whether I should pivot, add experiments, or confirm the result”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
result-to-claim.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip result-to-claim.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip result-to-claim.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Point the skill at a specific claim or W&B run ID so it judges one clear question rather than everything at once.
- Keep research_contract.md up to date, since the verdict is only as good as the stated intended claim.
- Run it once after experiments finish rather than scheduling it, as the skill itself advises.
What's inside
- SKILL.md
Redistributed unmodified from wanshuiyin/Auto-claude-code-research-in-sleep under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.