GrantMaster

Problem

Research grant applications often stall at concrete failures: unclear policy alignment, weak preliminary evidence, inconsistent proposal sections, superficial self-review, and untargeted defense preparation. GrantMaster breaks this into staged workflows instead of producing a single generic text.

How It Works

• Discipline adaptation: starts with education and can switch to medicine, engineering, agriculture, arts/sports, or basic science, loading matching policy sources and hot topics.
• Lifecycle coverage: handles policy scanning, topic diagnosis, literature search, proposal drafting, quality review, defense simulation, and deliverable compilation. State is saved to grant-master-state.json to reduce context loss across sessions.
• Knowledge source coordination: prefers personal knowledge in ima-mcp, falls back to web search, and can query a local library via the zotero skill without modifying entries.
• Review and output: provides fast, standard, and deep modes that check authenticity, feasibility, AI traces, compliance, and policy fit; output can target Markdown, JSON, or GB/T 9704 document formatting.

Limits

It is a research writing and QA assistant, not a reviewer. Scores are text-based estimates, deep mode consumes more tokens, and users should confirm discipline, level, and scope before formal submission.