yuema137/scientific-eval-environments
Skill Claude CodeCodex
Run a targeted historical coverage backfill for one canonical domain of this knowledge base — broad multi-source discovery, parallel primary-source deep review, Topic/Domain/Activity integration, Chinese mirror with independent review, and one reviewable PR. Use when a domain page is thin and the maintainer asks to…
yuema137/scientific-eval-environments
Instructions file
Claude Code instructions for yuema137/scientific-eval-environments, covering claude.md, layer model, canonical topic taxonomy (fixed), explanation style and first-appearance dates.
yuema137/scientific-eval-environments
Agent
You adversarially review one bilingual monthly report pair and revise both files in place when the report overclaims, contradicts itself, or leaves a structural promise unresolved.
yuema137/scientific-eval-environments
Agent
Independently revise one generated Chinese monthly report while preserving its facts and links.
yuema137/scientific-eval-environments
Agent
Translate the canonical English monthly report into a natural Chinese report.
yuema137/scientific-eval-environments
Agent
You write the canonical English monthly report for this repository.
yuema137/scientific-eval-environments
Agent
Phase 3 — assign accepted works to canonical Research Activities and update activity pages. Runs only if the Activity axis exists.
yuema137/scientific-eval-environments
Agent
Phase 4 — translate a group of changed English knowledge pages into their Chinese mirror, following repository bilingual conventions.
yuema137/scientific-eval-environments
Agent
Phase 5 — independently edit the Phase-4 Chinese translations for natural, idiomatic scientific Chinese. Distinct from the translator.
yuema137/scientific-eval-environments
Agent
Phase 3 — assign accepted works to canonical scientific/engineering Domains and update domain pages (one-way mapping).
yuema137/scientific-eval-environments
Agent
Optional final read-only sanity auditor over the staged update diff. Advisory only — deterministic validators and the phase gate are authoritative.
yuema137/scientific-eval-environments
Agent
Phase-1 metadata-only relevance triage. Scores a batch of candidates on whether they deserve expensive Phase-2 primary-source review. Reads only lightweight metadata.
yuema137/scientific-eval-environments
Agent
Phase 3 — assign accepted works to canonical Topics and update topic pages. Does not edit work-card files.
yuema137/scientific-eval-environments
Agent
Phase 2 — deeply review one candidate work from primary sources and either reject it or write a complete factual English work card.