Auto Paper Improvement Loop

Run controlled, multi-round review → implement → recompile cycles on a paper draft. Each review round uses a fresh context to prevent confirmation bias; an edit-whitelist gates what may be changed; state is checkpointed after each round so sessions can resume.

Use this skill when:

• a paper draft needs iterative writing quality improvement beyond a single-pass consistency edit
• reviewer independence matters: prior-context reviews inflate scores and miss real problems
• certain parts of the paper (theorems, numerics, citations) should be frozen during a writing pass
• a long improvement session may span multiple agent sessions and needs crash recovery
• you want a logged diff of what changed between draft versions

Do not use this skill as a substitute for real reviewer feedback — use paper-reviewer-simulator first to identify structural risks. Do not use this skill to make decisions about experimental results — use result-diagnosis or research-results-auditor before running improvement loops.

Pair this skill with: