Loading ToolBench
The Medical Writer's AI Toolkit — 33 expert prompts as callable MCP tools for pharmaceutical medical writing, curated by a CMPP-certified professional with a PhD and 10+ years of experience.
Medical Writer's AI Toolkit exposes 10 tools with consistent naming (all start with 'get_'), clear descriptions (100-200 chars, well-structured), and complete input schemas with typed parameters. However, output schemas are entirely undocumented, tools return plain strings with no structured format guidance. Error handling is minimal (only data-safety warnings in descriptions, no recovery guidance). Parameter descriptions are adequate but lack constraints (enums, ranges, formats). No tool annotations (readOnlyHint, destructiveHint) despite all tools being read-only. Composition is sound, each tool has one responsibility, but the lack of output documentation and error guidance limits LLM reasoning and recovery.
[PRO] Rewrite an abstract for a different audience while preserving scientific accuracy. Useful for patient summaries, nurse education, or general medical audiences. DATA SAFETY: Only paste published or approved text.
[FREE] Generate a prompt to draft a professional author invitation email. ICMJE authorship criteria aligned. Tone: professional and respectful of their time. Target length: 200-250 words.
[PRO] Draft a manuscript Introduction/Background section. Covers disease burden, treatment landscape, study rationale, and objective statement. AMA style. No references included.
[FREE] Generate a prompt to draft a LinkedIn post announcing a publication. Professional but engaging tone. Includes call to action and hashtag suggestions. Target length: 150-200 words.
[FREE] Generate a prompt to create a detailed manuscript outline. Includes section headers, word counts, key data points, and journal formatting notes. Suitable for original research, reviews, and case reports.
No output schema documentation. All tools return plain strings with no structured format guidance. LLMs cannot plan downstream processing or extract specific fields.
Missing tool annotations. All tools are read-only (return prompts, not state changes), but lack readOnlyHint annotation. Agents cannot distinguish safe tools from destructive ones without explicit hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
[PRO] Draft a manuscript Methods section. CONSORT aligned for RCTs, STROBE for observational studies. Covers design, eligibility, treatment arms, endpoints, and statistical analysis.
[FREE] Generate a prompt to draft a professional point-by-point reviewer response. Produces respectful, concise, data-driven responses that maximize acceptance chances. DATA SAFETY: Only reference published or approved data in supporting_data.
[PRO] Rewrite an abstract conclusion to be data-anchored and clinically meaningful. Addresses the #1 reason abstracts get rejected: weak conclusions. DATA SAFETY: Only use published or approved text.
[FREE] Generate a prompt to draft a structured journal abstract. Covers Background, Objective, Methods, Results, and Conclusions. CONSORT/STROBE aligned. Suitable for clinical manuscripts. DATA SAFETY: Only use published or approved summary data in the fields.
[PRO] Convert a structured abstract to a flowing unstructured paragraph. Required by many basic science journals (Nature, Science family). DATA SAFETY: Only paste published or approved text.
Parameter constraints missing. No enums, ranges, or format specifications. E.g., 'word_limit' (integer) lacks min/max bounds; 'reading_level' (string) lacks enum values; 'manuscript_type' lacks valid options. LLMs will hallucinate invalid values.
No error handling or recovery guidance. Data-safety warnings appear in descriptions but no actionable error messages or recovery paths. If an LLM passes unpublished data, the tool silently accepts it.
Inconsistent parameter naming. 'word_limit' vs 'word_count' vs 'word_length' across tools. 'target_journal' vs 'journal'. LLMs conflate similar names and may pass wrong values.