Rt Simulated Referee
Testing & Qualityrt-simulated-refereeRt Simulated Referee is an Agent Skill for testing & quality published in the brycewang-stanford/Awesome-Journal-Skills repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Use to rehearse peer review before submitting — a calibrated Associate Editor desk-screen plus 2–3 distinct-lens referees for the target venue, adversarially verified and synthesized into a referee report, a decision band, and a prioritized, skill-mapped fix list. A rehearsal that predicts the attack surface; it does not replace real review.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/brycewang-stanford/Awesome-Journal-Skills/blob/HEAD/Research-Toolkit-Skills/skills/rt-simulated-referee/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/rt-simulated-referee, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Simulated Referee (rt-simulated-referee)
Run the substantive gauntlet before a real referee does. Full protocol + a worked example:
shared-resources/submission-readiness/simulated-referee.md.
When to trigger
- After
rt-submission-readinessclears the mechanical bar. - When the author wants to know the decisive objection before submitting.
Calibrate first (this is what makes it realistic)
Pull the venue's bar from the target pack, not generic intuition: acceptance/desk-reject
reality + mechanics from resources/official-source-map.md; what this venue's referees
attack from *-referee-strategy + the shared
reviewer-objection-checklist;
the fit/identification/robustness bar from the pack's skills. Set strictness to the
venue's selectivity.
Protocol
- AE desk-screen — fit, general interest, fatal flaws → desk-reject risk; stop if a clear desk reject.
- Independent referees (distinct lenses: identification skeptic · contribution/fit · robustness/reproducibility) each write a structured report.
- Adversarial verification — every major concern must be real + specific + addressable or be downgraded.
- AE synthesis — one decision band (reject / major / minor / lean-accept) + the 1–3 decisive issues.
- Skill-mapped fix list — each decisive issue → the pack skill (and its execution bridge) that fixes it.
Orchestration
Spawn one subagent per role with the same manuscript + the calibrated venue bar; collect independent reports; a final AE-synthesis agent. Distinct lenses + independence catch more than a single read.
Hard rules
- Calibrate from the pack + source-map, never generic.
- Verify before reporting (real / specific / addressable).
- Map every decisive issue to a skill + execution bridge.
- It's a rehearsal — output a decision band and the decisive issues, never a fake accept probability.
Output format
【Venue】… (calibrated)
【Desk-screen】send out / desk reject — reason; risk low/med/high
【Decision band】reject / major / minor / lean-accept
【Decisive issues (1–3)】each with owning pack skill + concrete fix
【Author action plan】ordered; tie each to a skill + tool
Next: address the decisive issues, then rt-response-to-referees for a real R&R.