sk Skill · zjunlp
auto-iteration-loop
Autonomous research review loop that consumes /auto-verify's four-state output (PASS / FAIL / INCONCLUSIVE / ZERO_ELIGIBLE_VARIANTS / deferred) and routes each claim to the right back-edge — brief audit, two-phase FAIL handling (variant-integrity fix then optional claim-stage re-entry), main-experiment-script fix, or variant-only fix — under a unified iteration budget. Configure the reviewer LLM via llm-chat MCP…
Open on skills.sh ↗read 2026-09-15
- installs 8w
- 0
- 30-day movement
- starts with the next reading
- Related entries
- 1
- Connections
- 12
jsonmarkdownbashPython
- Host repository
- zjunlp/Mechanist
- Allowed tools
- Bash(*), Read, Grep, Glob, Write, Edit, WebSearch, WebFetch, Agent, AskUserQuestion, Skill, mcp__llm-chat__chat
- Host stars
- 75
- Host language
- Python